diff --git a/.travis.yml b/.travis.yml
index 9bf0c215c..d44a93d84 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -46,8 +46,8 @@ matrix:
         - cd ../..
 
         - python src/plasma/test/test.py valgrind
-
         - python src/photon/test/test.py valgrind
+        - python src/global_scheduler/test/test.py valgrind
 
 install:
   - ./install-dependencies.sh
@@ -57,10 +57,6 @@ install:
   - sudo python setup.py install
   - cd ../../../..
 
-  - cd src/photon
-  - sudo python setup.py install
-  - cd ../..
-
   - cd lib/python
   - sudo python setup.py install
   - cd ../..
@@ -69,6 +65,7 @@ script:
   - python src/common/test/test.py
   - python src/plasma/test/test.py
   - python src/photon/test/test.py
+  - python src/global_scheduler/test/test.py
 
   - python test/runtest.py
   - python test/array_test.py
diff --git a/build-webui.sh b/build-webui.sh
index 1c7a726ef..07ef7eae6 100755
--- a/build-webui.sh
+++ b/build-webui.sh
@@ -1,5 +1,8 @@
 #!/usr/bin/env bash
 
+# Cause the script to exit if a single command fails.
+set -e
+
 ROOT_DIR=$(cd "$(dirname "${BASH_SOURCE:-$0}")"; pwd)
 
 unamestr="$(uname)"
diff --git a/build.sh b/build.sh
index 5fe6559d6..9b87f55c9 100755
--- a/build.sh
+++ b/build.sh
@@ -19,19 +19,23 @@ fi
 COMMON_DIR="$ROOT_DIR/src/common"
 PLASMA_DIR="$ROOT_DIR/src/plasma"
 PHOTON_DIR="$ROOT_DIR/src/photon"
+GLOBAL_SCHEDULER_DIR="$ROOT_DIR/src/global_scheduler"
 
 PYTHON_DIR="$ROOT_DIR/lib/python"
 PYTHON_COMMON_DIR="$PYTHON_DIR/common"
 PYTHON_PLASMA_DIR="$PYTHON_DIR/plasma"
 PYTHON_PHOTON_DIR="$PYTHON_DIR/photon"
+PYTHON_GLOBAL_SCHEDULER_DIR="$PYTHON_DIR/global_scheduler"
 
 pushd "$COMMON_DIR"
+  make clean
   make
   make test
 popd
 cp "$COMMON_DIR/thirdparty/redis-3.2.3/src/redis-server" "$PYTHON_COMMON_DIR/thirdparty/redis-3.2.3/src/"
 
 pushd "$PLASMA_DIR"
+  make clean
   make
   make test
   pushd "$PLASMA_DIR/build"
@@ -45,6 +49,7 @@ cp "$PLASMA_DIR/lib/python/plasma.py" "$PYTHON_PLASMA_DIR/lib/python/"
 cp "$PLASMA_DIR/lib/python/libplasma.so" "$PYTHON_PLASMA_DIR/lib/python/"
 
 pushd "$PHOTON_DIR"
+  make clean
   make
   pushd "$PHOTON_DIR/build"
     cmake ..
@@ -52,4 +57,12 @@ pushd "$PHOTON_DIR"
   popd
 popd
 cp "$PHOTON_DIR/build/photon_scheduler" "$PYTHON_PHOTON_DIR/build/"
-cp "$PHOTON_DIR/photon/libphoton.so" "$PYTHON_PHOTON_DIR/"
+cp "$PHOTON_DIR/photon/libphoton.so" "$PYTHON_PHOTON_DIR/photon/"
+cp "$PHOTON_DIR/photon/photon_services.py" "$PYTHON_PHOTON_DIR/photon/"
+
+pushd "$GLOBAL_SCHEDULER_DIR"
+  make clean
+  make
+popd
+cp "$GLOBAL_SCHEDULER_DIR/build/global_scheduler" "$PYTHON_GLOBAL_SCHEDULER_DIR/build/"
+cp "$GLOBAL_SCHEDULER_DIR/lib/python/global_scheduler_services.py" "$PYTHON_GLOBAL_SCHEDULER_DIR/lib/python/"
diff --git a/lib/python/global_scheduler/__init__.py b/lib/python/global_scheduler/__init__.py
new file mode 100644
index 000000000..c92c0e1b6
--- /dev/null
+++ b/lib/python/global_scheduler/__init__.py
@@ -0,0 +1 @@
+from lib.python.global_scheduler_services import *
diff --git a/lib/python/global_scheduler/build/.gitkeep b/lib/python/global_scheduler/build/.gitkeep
new file mode 100644
index 000000000..e69de29bb
diff --git a/lib/python/global_scheduler/lib/__init__.py b/lib/python/global_scheduler/lib/__init__.py
new file mode 100644
index 000000000..e69de29bb
diff --git a/lib/python/global_scheduler/lib/python/__init__.py b/lib/python/global_scheduler/lib/python/__init__.py
new file mode 100644
index 000000000..e69de29bb
diff --git a/lib/python/photon/__init__.py b/lib/python/photon/__init__.py
index 27f7c578f..73b874cd4 100644
--- a/lib/python/photon/__init__.py
+++ b/lib/python/photon/__init__.py
@@ -1 +1 @@
-from libphoton import *
+from photon import *
diff --git a/lib/python/photon/photon/__init__.py b/lib/python/photon/photon/__init__.py
new file mode 100644
index 000000000..d34d8d3e1
--- /dev/null
+++ b/lib/python/photon/photon/__init__.py
@@ -0,0 +1,2 @@
+from libphoton import *
+from photon_services import *
diff --git a/lib/python/ray/services.py b/lib/python/ray/services.py
index 8d92e502e..e3eaa7c3f 100644
--- a/lib/python/ray/services.py
+++ b/lib/python/ray/services.py
@@ -4,14 +4,16 @@ import psutil
 import os
 import random
 import signal
-import sys
-import subprocess
 import string
+import subprocess
+import sys
 import time
 
 # Ray modules
 import config
+import photon
 import plasma
+import global_scheduler
 
 # all_processes is a list of the scheduler, object store, and worker processes
 # that have been started by this services module if Ray is being used in local
@@ -96,14 +98,33 @@ def start_redis(num_retries=20, cleanup=True):
     counter += 1
   raise Exception("Couldn't start Redis.")
 
+def start_global_scheduler(redis_address, cleanup=True):
+  """Start a global scheduler process.
+
+  Args:
+    redis_address (str): The address of the Redis instance.
+    cleanup (bool): True if using Ray in local mode. If cleanup is true, then
+      this process will be killed by serices.cleanup() when the Python process
+      that imported services exits.
+  """
+  p = global_scheduler.start_global_scheduler(redis_address)
+  if cleanup:
+    all_processes.append(p)
+
 def start_local_scheduler(redis_address, plasma_store_name, cleanup=True):
-  local_scheduler_filepath = os.path.join(os.path.dirname(os.path.abspath(__file__)), "../photon/build/photon_scheduler")
-  if RUN_PHOTON_PROFILER:
-    local_scheduler_prefix = ["valgrind", "--tool=callgrind", local_scheduler_filepath]
-  else:
-    local_scheduler_prefix = [local_scheduler_filepath]
-  local_scheduler_name = "/tmp/scheduler{}".format(random_name())
-  p = subprocess.Popen(local_scheduler_prefix + ["-s", local_scheduler_name, "-r", redis_address, "-p", plasma_store_name])
+  """Start a local scheduler process.
+
+  Args:
+    redis_address (str): The address of the Redis instance.
+    plasma_store_name (str): The name of the plasma store socket to connect to.
+    cleanup (bool): True if using Ray in local mode. If cleanup is true, then
+      this process will be killed by serices.cleanup() when the Python process
+      that imported services exits.
+
+  Return:
+    The name of the local scheduler socket.
+  """
+  local_scheduler_name, p = photon.start_local_scheduler(plasma_store_name, redis_address=redis_address, use_profiler=RUN_PHOTON_PROFILER)
   if cleanup:
     all_processes.append(p)
   return local_scheduler_name
@@ -113,29 +134,27 @@ def start_objstore(node_ip_address, redis_address, cleanup=True):
 
   Args:
     node_ip_address (str): The ip address of the node running the object store.
+    redis_address (str): The address of the Redis instance to connect to.
     cleanup (bool): True if using Ray in local mode. If cleanup is true, then
       this process will be killed by serices.cleanup() when the Python process
       that imported services exits.
+
+  Return:
+    A tuple of the Plasma store socket name, the Plasma manager socket name, and
+      the plasma manager port.
   """
-  # Let the object store use a fraction of the system memory.
+  # Compute a fraction of the system memory for the Plasma store to use.
   system_memory = psutil.virtual_memory().total
   plasma_store_memory = int(system_memory * 0.75)
-  plasma_store_filepath = os.path.join(os.path.abspath(os.path.dirname(__file__)), "../plasma/build/plasma_store")
-  if RUN_PLASMA_STORE_PROFILER:
-    plasma_store_prefix = ["valgrind", "--tool=callgrind", plasma_store_filepath]
-  else:
-    plasma_store_prefix = [plasma_store_filepath]
-  store_name = "/tmp/ray_plasma_store{}".format(random_name())
-  p1 = subprocess.Popen(plasma_store_prefix + ["-s", store_name, "-m", str(plasma_store_memory)])
-
-  manager_name = "/tmp/ray_plasma_manager{}".format(random_name())
-  p2, manager_port = plasma.start_plasma_manager(store_name, manager_name, redis_address, run_profiler=RUN_PLASMA_MANAGER_PROFILER)
-
+  # Start the Plasma store.
+  plasma_store_name, p1 = plasma.start_plasma_store(plasma_store_memory=plasma_store_memory, use_profiler=RUN_PLASMA_STORE_PROFILER)
+  # Start the plasma manager.
+  plasma_manager_name, p2, plasma_manager_port = plasma.start_plasma_manager(plasma_store_name, redis_address, run_profiler=RUN_PLASMA_MANAGER_PROFILER)
   if cleanup:
     all_processes.append(p1)
     all_processes.append(p2)
 
-  return store_name, manager_name, manager_port
+  return plasma_store_name, plasma_manager_name, plasma_manager_port
 
 def start_worker(address_info, worker_path, cleanup=True):
   """This method starts a worker process.
@@ -186,8 +205,8 @@ def start_ray_local(node_ip_address="127.0.0.1", num_workers=0, worker_path=None
       worker.
 
   Returns:
-    This returns a tuple of three things. The first element is a tuple of the
-    Redis hostname and port. The second
+    This returns a dictionary of the address information for the processes that
+      were started.
   """
   if worker_path is None:
     worker_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), "workers/default_worker.py")
@@ -195,12 +214,14 @@ def start_ray_local(node_ip_address="127.0.0.1", num_workers=0, worker_path=None
   redis_port = start_redis(cleanup=True)
   redis_address = address(node_ip_address, redis_port)
   time.sleep(0.1)
+  # Start the global scheduler.
+  start_global_scheduler(redis_address, cleanup=True)
   # Start Plasma.
   object_store_name, object_store_manager_name, object_store_manager_port = start_objstore(node_ip_address, redis_address, cleanup=True)
-  # Start the local scheduler.
   time.sleep(0.1)
+  # Start the local scheduler.
   local_scheduler_name = start_local_scheduler(redis_address, object_store_name, cleanup=True)
-  time.sleep(0.2)
+  time.sleep(0.1)
   # Aggregate the address information together.
   address_info = {"node_ip_address": node_ip_address,
                   "redis_port": redis_port,
@@ -210,7 +231,6 @@ def start_ray_local(node_ip_address="127.0.0.1", num_workers=0, worker_path=None
   # Start the workers.
   for _ in range(num_workers):
     start_worker(address_info, worker_path, cleanup=True)
-  time.sleep(0.3)
   # Return the addresses of the relevant processes.
   start_webui(redis_port)
   return address_info
diff --git a/lib/python/ray/worker.py b/lib/python/ray/worker.py
index 7f304cce3..160fd18ca 100644
--- a/lib/python/ray/worker.py
+++ b/lib/python/ray/worker.py
@@ -643,8 +643,7 @@ def init(start_ray_local=False, num_workers=None, driver_mode=SCRIPT_MODE):
   # corresponing call to disconnect will happen in the call to cleanup() when
   # the Python script exits.
   connect(address_info, driver_mode, worker=global_worker)
-  if driver_mode != PYTHON_MODE:
-    return "{}:{}".format(address_info["node_ip_address"], address_info["redis_port"])
+  return address_info
 
 def cleanup(worker=global_worker):
   """Disconnect the driver, and terminate any processes started in init.
diff --git a/lib/python/setup.py b/lib/python/setup.py
index e1dc92b76..7d5c444eb 100644
--- a/lib/python/setup.py
+++ b/lib/python/setup.py
@@ -26,7 +26,8 @@ setup(name="ray",
                                "build/plasma_manager",
                                "lib/python/libplasma.so"],
                     "photon": ["build/photon_scheduler",
-                               "libphoton.so"]},
+                               "photon/libphoton.so"],
+                    "global_scheduler": ["build/global_scheduler"]},
       data_files=datafiles,
       cmdclass={"install": install},
       install_requires=["numpy",
diff --git a/src/common/Makefile b/src/common/Makefile
index 9a7d670f5..a19ac1f19 100644
--- a/src/common/Makefile
+++ b/src/common/Makefile
@@ -4,7 +4,7 @@ BUILD = build
 
 all: hiredis $(BUILD)/libcommon.a
 
-$(BUILD)/libcommon.a: event_loop.o common.o task.o io.o state/redis.o state/table.o state/object_table.o state/task_table.o thirdparty/ae/ae.o thirdparty/sha256.o
+$(BUILD)/libcommon.a: event_loop.o common.o task.o io.o net.o state/redis.o state/table.o state/object_table.o state/task_table.o state/db_client_table.o thirdparty/ae/ae.o thirdparty/sha256.o
 	ar rcs $@ $^
 
 $(BUILD)/common_tests: test/common_tests.c $(BUILD)/libcommon.a
diff --git a/src/common/common.c b/src/common/common.c
index 2b321c94d..c5450db3e 100644
--- a/src/common/common.c
+++ b/src/common/common.c
@@ -25,13 +25,17 @@ unique_id globally_unique_id(void) {
 }
 
 bool object_ids_equal(object_id first_id, object_id second_id) {
-  return UNIQUE_ID_EQ(first_id, second_id) ? true : false;
+  return UNIQUE_ID_EQ(first_id, second_id);
 }
 
 bool object_id_is_nil(object_id id) {
   return object_ids_equal(id, NIL_OBJECT_ID);
 }
 
+bool db_client_ids_equal(db_client_id first_id, db_client_id second_id) {
+  return UNIQUE_ID_EQ(first_id, second_id);
+}
+
 char *sha1_to_hex(const unsigned char *sha1, char *buffer) {
   static const char hex[] = "0123456789abcdef";
   char *buf = buffer;
diff --git a/src/common/common.h b/src/common/common.h
index 12ec53e83..9073a9b67 100644
--- a/src/common/common.h
+++ b/src/common/common.h
@@ -129,4 +129,15 @@ bool object_ids_equal(object_id first_id, object_id second_id);
  */
 bool object_id_is_nil(object_id id);
 
+typedef unique_id db_client_id;
+
+/**
+ * Compare two db client IDs.
+ *
+ * @param first_id The first db client ID to compare.
+ * @param second_id The first db client ID to compare.
+ * @return True if the db client IDs are the same and false otherwise.
+ */
+bool db_client_ids_equal(db_client_id first_id, db_client_id second_id);
+
 #endif
diff --git a/src/common/logging.c b/src/common/logging.c
index 09cad7856..100b2777d 100644
--- a/src/common/logging.c
+++ b/src/common/logging.c
@@ -10,7 +10,7 @@
 
 static const char *log_levels[5] = {"DEBUG", "INFO", "WARN", "ERROR", "FATAL"};
 static const char *log_fmt =
-    "HMSET log:%s:%s:%s log_level %s event_type %s message %s timestamp %s";
+    "HMSET log:%s:%s log_level %s event_type %s message %s timestamp %s";
 
 struct ray_logger_impl {
   /* String that identifies this client type. */
@@ -56,25 +56,23 @@ void ray_log(ray_logger *logger,
   gettimeofday(&tv, NULL);
   utstring_printf(timestamp, "%ld.%ld", tv.tv_sec, (long) tv.tv_usec);
 
-  UT_string *origin_id;
-  utstring_new(origin_id);
+  UT_string *formatted_message;
+  utstring_new(formatted_message);
+  /* Fill out everything except the client ID, which is binary data. */
+  utstring_printf(formatted_message, log_fmt, utstring_body(timestamp), "%b",
+                  log_levels[log_level], event_type, message,
+                  utstring_body(timestamp));
   if (logger->is_direct) {
     db_handle *db = (db_handle *) logger->conn;
-    utstring_printf(origin_id, "%" PRId64 ":%s", db->client_id, "");
-    redisAsyncCommand(db->context, NULL, NULL, log_fmt,
-                      utstring_body(timestamp), logger->client_type,
-                      utstring_body(origin_id), log_levels[log_level],
-                      event_type, message, utstring_body(timestamp));
+    /* Fill in the client ID and send the message to Redis. */
+    redisAsyncCommand(db->context, NULL, NULL, utstring_body(formatted_message),
+                      (char *) db->client.id, sizeof(db_client_id));
   } else {
     /* If we don't own a Redis connection, we leave our client
      * ID to be filled in by someone else. */
-    utstring_printf(origin_id, "%s:%s", "%ld", "%ld");
     int *socket_fd = (int *) logger->conn;
-    write_formatted_log_message(*socket_fd, log_fmt, utstring_body(timestamp),
-                                logger->client_type, utstring_body(origin_id),
-                                log_levels[log_level], event_type, message,
-                                utstring_body(timestamp));
+    write_log_message(*socket_fd, utstring_body(formatted_message));
   }
-  utstring_free(origin_id);
+  utstring_free(formatted_message);
   utstring_free(timestamp);
 }
diff --git a/src/common/net.c b/src/common/net.c
new file mode 100644
index 000000000..3e56cff50
--- /dev/null
+++ b/src/common/net.c
@@ -0,0 +1,13 @@
+#include "net.h"
+
+#include "common.h"
+
+int parse_ip_addr_port(const char *ip_addr_port, char *ip_addr, int *port) {
+  char port_str[6];
+  int parsed = sscanf(ip_addr_port, "%15[0-9.]:%5[0-9]", ip_addr, port_str);
+  if (parsed != 2) {
+    return -1;
+  }
+  *port = atoi(port_str);
+  return 0;
+}
diff --git a/src/common/net.h b/src/common/net.h
new file mode 100644
index 000000000..109cdf3fa
--- /dev/null
+++ b/src/common/net.h
@@ -0,0 +1,9 @@
+#ifndef NET_H
+#define NET_H
+
+/* Helper function to parse a string of the form <IP address>:<port> into the
+ * given ip_addr and port pointers. The ip_addr buffer must already be
+ * allocated. Return 0 upon success and -1 upon failure. */
+int parse_ip_addr_port(const char *ip_addr_port, char *ip_addr, int *port);
+
+#endif /* NET_H */
diff --git a/src/common/state/db.h b/src/common/state/db.h
index 96ecf1e79..e50adbbd8 100644
--- a/src/common/state/db.h
+++ b/src/common/state/db.h
@@ -1,32 +1,55 @@
 #ifndef DB_H
 #define DB_H
 
+#include "common.h"
 #include "event_loop.h"
 
 typedef struct db_handle db_handle;
 
-/* Connect to the global system store at address and port. Returns
- * a handle to the database, which must be freed with db_disconnect
- * after use. */
+/**
+ * Connect to the global system store.
+ *
+ * @param db_address The hostname to use to connect to the database.
+ * @param db_port The port to use to connect to the database.
+ * @param client_type The type of this client.
+ * @param client_addr The hostname of the client that is connecting. If not
+ *        relevant, set this to the empty string.
+ * @param client_port The port of the client that is connecting. If not
+ *        relevant, set this to -1.
+ * @return This returns a handle to the database, which must be freed with
+ *         db_disconnect after use.
+ */
+
 db_handle *db_connect(const char *db_address,
                       int db_port,
                       const char *client_type,
                       const char *client_addr,
                       int client_port);
 
-/* Attach global system store connection to event loop. */
+/**
+ * Attach global system store connection to an event loop. Callbacks from
+ * queries to the global system store will trigger events in the event loop.
+ *
+ * @param db The database in question.
+ * @param loop The event loop to attach to.
+ * @return Void.
+ */
 void db_attach(db_handle *db, event_loop *loop);
 
-/* Disconnect from the global system store. */
+/**
+ * Disconnect from the global system store.
+ *
+ * @param db The database connection to close and clean up.
+ * @return Void.
+ */
 void db_disconnect(db_handle *db);
 
 /**
- * Returns the client ID, according to the database.
+ * Returns the db client ID.
  *
  * @param db The handle to the database.
- * @returns int The client ID for this connection to the database. If
- *          this client has no connection to the database, returns -1.
+ * @returns int The db client ID for this connection to the database.
  */
-int get_client_id(db_handle *db);
+db_client_id get_db_client_id(db_handle *db);
 
 #endif
diff --git a/src/common/state/db_client_table.c b/src/common/state/db_client_table.c
new file mode 100644
index 000000000..5a706c756
--- /dev/null
+++ b/src/common/state/db_client_table.c
@@ -0,0 +1,20 @@
+#include "db_client_table.h"
+#include "redis.h"
+
+void db_client_table_subscribe(
+    db_handle *db_handle,
+    db_client_table_subscribe_callback subscribe_callback,
+    void *subscribe_context,
+    retry_info *retry,
+    db_client_table_done_callback done_callback,
+    void *user_context) {
+  db_client_table_subscribe_data *sub_data =
+      malloc(sizeof(db_client_table_subscribe_data));
+  utarray_push_back(db_handle->callback_freelist, &sub_data);
+  sub_data->subscribe_callback = subscribe_callback;
+  sub_data->subscribe_context = subscribe_context;
+
+  init_table_callback(db_handle, NIL_ID, __func__, sub_data, retry,
+                      done_callback, redis_db_client_table_subscribe,
+                      user_context);
+}
diff --git a/src/common/state/db_client_table.h b/src/common/state/db_client_table.h
new file mode 100644
index 000000000..3ea3949a3
--- /dev/null
+++ b/src/common/state/db_client_table.h
@@ -0,0 +1,48 @@
+#ifndef DB_CLIENT_TABLE_H
+#define DB_CLIENT_TABLE_H
+
+#include "db.h"
+#include "table.h"
+
+typedef void (*db_client_table_done_callback)(db_client_id db_client_id,
+                                              void *user_context);
+
+/*
+ *  ==== Subscribing to the db client table ====
+ */
+
+/* Callback for subscribing to the db client table. */
+typedef void (*db_client_table_subscribe_callback)(db_client_id db_client_id,
+                                                   const char *client_type,
+                                                   void *user_context);
+
+/**
+ * Register a callback for a db client table event.
+ *
+ * @param db_handle Database handle.
+ * @param subscribe_callback Callback that will be called when the db client
+ *        table is updated.
+ * @param subscribe_context Context that will be passed into the
+ *        subscribe_callback.
+ * @param retry Information about retrying the request to the database.
+ * @param done_callback Function to be called when database returns result.
+ * @param user_context Data that will be passed to done_callback and
+ *        fail_callback.
+ * @return Void.
+ */
+void db_client_table_subscribe(
+    db_handle *db_handle,
+    db_client_table_subscribe_callback subscribe_callback,
+    void *subscribe_context,
+    retry_info *retry,
+    db_client_table_done_callback done_callback,
+    void *user_context);
+
+/* Data that is needed to register db client table subscribe callbacks with the
+ * state database. */
+typedef struct {
+  db_client_table_subscribe_callback subscribe_callback;
+  void *subscribe_context;
+} db_client_table_subscribe_data;
+
+#endif /* DB_CLIENT_TABLE_H */
diff --git a/src/common/state/redis.c b/src/common/state/redis.c
index da60f08f1..4ea8753e4 100644
--- a/src/common/state/redis.c
+++ b/src/common/state/redis.c
@@ -9,6 +9,7 @@
 
 #include "common.h"
 #include "db.h"
+#include "db_client_table.h"
 #include "object_table.h"
 #include "task.h"
 #include "task_table.h"
@@ -54,21 +55,26 @@ db_handle *db_connect(const char *address,
   db_handle *db = malloc(sizeof(db_handle));
   /* Sync connection for initial handshake */
   redisReply *reply;
-  long long num_clients;
   redisContext *context = redisConnect(address, port);
   CHECK_REDIS_CONNECT(redisContext, context, "could not connect to redis %s:%d",
                       address, port);
   /* Add new client using optimistic locking. */
+  db_client_id client = globally_unique_id();
   while (1) {
     reply = redisCommand(context, "WATCH %s", client_type);
     freeReplyObject(reply);
     reply = redisCommand(context, "HLEN %s", client_type);
-    num_clients = reply->integer;
     freeReplyObject(reply);
     reply = redisCommand(context, "MULTI");
     freeReplyObject(reply);
-    reply = redisCommand(context, "HSET %s %lld %s:%d", client_type,
-                         num_clients, client_addr, client_port);
+    reply = redisCommand(
+        context,
+        "HMSET db_clients:%b client_type %s address %s:%d db_client_id %b",
+        (char *) client.id, sizeof(db_client_id), client_type, client_addr,
+        client_port, (char *) client.id, sizeof(db_client_id));
+    freeReplyObject(reply);
+    reply = redisCommand(context, "PUBLISH db_clients %b:%s",
+                         (char *) client.id, sizeof(db_client_id), client_type);
     freeReplyObject(reply);
     reply = redisCommand(context, "EXEC");
     CHECK(reply);
@@ -80,8 +86,8 @@ db_handle *db_connect(const char *address,
   }
 
   db->client_type = strdup(client_type);
-  db->client_id = num_clients;
-  db->service_cache = NULL;
+  db->client = client;
+  db->db_client_cache = NULL;
   db->sync_context = context;
   utarray_new(db->callback_freelist, &ut_ptr_icd);
 
@@ -103,10 +109,10 @@ void db_disconnect(db_handle *db) {
   redisFree(db->sync_context);
   redisAsyncFree(db->context);
   redisAsyncFree(db->sub_context);
-  service_cache_entry *e, *tmp;
-  HASH_ITER(hh, db->service_cache, e, tmp) {
+  db_client_cache_entry *e, *tmp;
+  HASH_ITER(hh, db->db_client_cache, e, tmp) {
     free(e->addr);
-    HASH_DEL(db->service_cache, e);
+    HASH_DELETE(hh, db->db_client_cache, e);
     free(e);
   }
   free(db->client_type);
@@ -200,10 +206,11 @@ void redis_object_table_add(table_callback_data *callback_data) {
   CHECK(callback_data);
   db_handle *db = callback_data->db_handle;
   object_id id = callback_data->id;
-  int status =
-      redisAsyncCommand(db->context, redis_object_table_add_callback,
-                        (void *) callback_data->timer_id, "SADD obj:%b %d",
-                        id.id, sizeof(object_id), db->client_id);
+  int status = redisAsyncCommand(db->context, redis_object_table_add_callback,
+                                 (void *) callback_data->timer_id,
+                                 "SADD obj:%b %b", id.id, sizeof(object_id),
+                                 (char *) db->client.id, sizeof(db_client_id));
+
   if ((status == REDIS_ERR) || db->context->err) {
     LOG_REDIS_DEBUG(db->context, "could not add object_table entry");
   }
@@ -330,21 +337,26 @@ void redis_result_table_lookup(table_callback_data *callback_data) {
  *
  * @param db The database handle.
  * @param index The index of the plasma manager.
- * @param *manager The pointer where the IP address of the manager gets written.
+ * @param manager The pointer where the IP address of the manager gets written.
  * @return Void.
  */
-void redis_get_cached_service(db_handle *db, int index, const char **manager) {
-  service_cache_entry *entry;
-  HASH_FIND_INT(db->service_cache, &index, entry);
+void redis_get_cached_db_client(db_handle *db,
+                                db_client_id db_client_id,
+                                const char **manager) {
+  db_client_cache_entry *entry;
+  HASH_FIND(hh, db->db_client_cache, &db_client_id, sizeof(db_client_id),
+            entry);
   if (!entry) {
-    /* This is a very rare case. */
+    /* This is a very rare case. It should happen at most once per db client. */
     redisReply *reply =
-        redisCommand(db->sync_context, "HGET %s %lld", db->client_type, index);
+        redisCommand(db->sync_context, "HGET db_clients:%b address",
+                     (char *) db_client_id.id, sizeof(db_client_id));
     CHECK(reply->type == REDIS_REPLY_STRING);
-    entry = malloc(sizeof(service_cache_entry));
-    entry->service_id = index;
+    entry = malloc(sizeof(db_client_cache_entry));
+    entry->db_client_id = db_client_id;
     entry->addr = strdup(reply->str);
-    HASH_ADD_INT(db->service_cache, service_id, entry);
+    HASH_ADD(hh, db->db_client_cache, db_client_id, sizeof(db_client_id),
+             entry);
     freeReplyObject(reply);
   }
   *manager = entry->addr;
@@ -356,15 +368,15 @@ void redis_object_table_get_entry(redisAsyncContext *c,
   REDIS_CALLBACK_HEADER(db, callback_data, r)
   redisReply *reply = r;
 
-  int *managers = malloc(reply->elements * sizeof(int));
+  db_client_id *managers = malloc(reply->elements * sizeof(db_client_id));
   int64_t manager_count = reply->elements;
 
   if (reply->type == REDIS_REPLY_ARRAY) {
     const char **manager_vector = malloc(manager_count * sizeof(char *));
-    for (int j = 0; j < reply->elements; j++) {
+    for (int j = 0; j < reply->elements; ++j) {
       CHECK(reply->element[j]->type == REDIS_REPLY_STRING);
-      managers[j] = atoi(reply->element[j]->str);
-      redis_get_cached_service(db, managers[j], manager_vector + j);
+      memcpy(managers[j].id, reply->element[j]->str, sizeof(db_client_id));
+      redis_get_cached_db_client(db, managers[j], manager_vector + j);
     }
 
     object_table_lookup_done_callback done_callback =
@@ -626,10 +638,62 @@ void redis_task_table_subscribe(table_callback_data *callback_data) {
   }
 }
 
-int get_client_id(db_handle *db) {
-  if (db) {
-    return db->client_id;
-  } else {
-    return -1;
+/*
+ *  ==== db client table callbacks ====
+ */
+
+void redis_db_client_table_subscribe_callback(redisAsyncContext *c,
+                                              void *r,
+                                              void *privdata) {
+  REDIS_CALLBACK_HEADER(db, callback_data, r)
+  redisReply *reply = r;
+
+  CHECK(reply->type == REDIS_REPLY_ARRAY);
+  /* If this condition is true, we got the initial message that acknowledged the
+   * subscription. */
+  CHECK(reply->elements > 2);
+  /* First entry is message type, then possibly the regex we psubscribed to,
+   * then topic, then payload. */
+  redisReply *payload = reply->element[reply->elements - 1];
+  if (payload->str == NULL) {
+    if (callback_data->done_callback) {
+      db_client_table_done_callback done_callback =
+          callback_data->done_callback;
+      done_callback(callback_data->id, callback_data->user_context);
+    }
+    /* Note that we do not destroy the callback data yet because the
+     * subscription callback needs this data. */
+    event_loop_remove_timer(db->loop, callback_data->timer_id);
+    return;
+  }
+  /* Otherwise, parse the payload and call the callback. */
+  db_client_table_subscribe_data *data = callback_data->data;
+  db_client_id client;
+  memcpy(client.id, payload->str, sizeof(db_client_id));
+  /* We subtract 1 + sizeof(db_client_id) to compute the length of the
+   * client_type string, and we add 1 to null-terminate the string. */
+  int client_type_length = payload->len - 1 - sizeof(db_client_id) + 1;
+  char *client_type = malloc(client_type_length);
+  memcpy(client_type, &payload->str[1 + sizeof(db_client_id)],
+         client_type_length);
+  if (data->subscribe_callback) {
+    data->subscribe_callback(client, client_type, data->subscribe_context);
+  }
+  free(client_type);
+}
+
+void redis_db_client_table_subscribe(table_callback_data *callback_data) {
+  db_handle *db = callback_data->db_handle;
+  int status = redisAsyncCommand(
+      db->sub_context, redis_db_client_table_subscribe_callback,
+      (void *) callback_data->timer_id, "SUBSCRIBE db_clients");
+  if ((status == REDIS_ERR) || db->sub_context->err) {
+    LOG_REDIS_DEBUG(db->sub_context,
+                    "error in db_client_table_register_callback");
   }
 }
+
+db_client_id get_db_client_id(db_handle *db) {
+  CHECK(db != NULL);
+  return db->client;
+}
diff --git a/src/common/state/redis.h b/src/common/state/redis.h
index 026cf30f9..0b4742df4 100644
--- a/src/common/state/redis.h
+++ b/src/common/state/redis.h
@@ -11,33 +11,33 @@
 #include "utarray.h"
 
 typedef struct {
-  /** Unique ID for this service. */
-  int service_id;
-  /** IP address and port of this service. */
+  /** Unique ID for this db client. */
+  db_client_id db_client_id;
+  /** IP address and port of this db client. */
   char *addr;
   /** Handle for the uthash table. */
   UT_hash_handle hh;
-} service_cache_entry;
+} db_client_cache_entry;
 
 struct db_handle {
   /** String that identifies this client type. */
   char *client_type;
-  /** Unique ID for this client within the type. */
-  int64_t client_id;
+  /** Unique ID for this client. */
+  db_client_id client;
   /** Redis context for this global state store connection. */
   redisAsyncContext *context;
-  /** Redis context for "subscribe" communication.
-   * Yes, we need a separate one for that, see
-   * https://github.com/redis/hiredis/issues/55 */
+  /** Redis context for "subscribe" communication. Yes, we need a separate one
+   *  for that, see https://github.com/redis/hiredis/issues/55. */
   redisAsyncContext *sub_context;
   /** The event loop this global state store connection is part of. */
   event_loop *loop;
   /** Index of the database connection in the event loop */
   int64_t db_index;
-  /** Cache for the IP addresses of services. */
-  service_cache_entry *service_cache;
-  /** Redis context for synchronous connections.
-   * Should only be used very rarely, it is not asynchronous. */
+  /** Cache for the IP addresses of db clients. This is a hash table mapping
+   *  client IDs to addresses. */
+  db_client_cache_entry *db_client_cache;
+  /** Redis context for synchronous connections. This should only be used very
+   *  rarely, it is not asynchronous. */
   redisContext *sync_context;
   /** Data structure for callbacks that needs to be freed. */
   UT_array *callback_freelist;
@@ -168,4 +168,13 @@ void redis_task_table_publish_publish_callback(redisAsyncContext *c,
  */
 void redis_task_table_subscribe(table_callback_data *callback_data);
 
+/**
+ * Subscribe to updates from the db client table.
+ *
+ * @param callback_data Data structure containing redis connection and timeout
+ *        information.
+ * @return Void.
+ */
+void redis_db_client_table_subscribe(table_callback_data *callback_data);
+
 #endif /* REDIS_H */
diff --git a/src/common/task.c b/src/common/task.c
index ea35c06f4..d7e8102dc 100644
--- a/src/common/task.c
+++ b/src/common/task.c
@@ -66,7 +66,7 @@ struct task_spec_impl {
    (ARGS_VALUE_SIZE))
 
 bool task_ids_equal(task_id first_id, task_id second_id) {
-  return UNIQUE_ID_EQ(first_id, second_id) ? true : false;
+  return UNIQUE_ID_EQ(first_id, second_id);
 }
 
 bool task_id_is_nil(task_id id) {
@@ -74,7 +74,7 @@ bool task_id_is_nil(task_id id) {
 }
 
 bool function_ids_equal(function_id first_id, function_id second_id) {
-  return UNIQUE_ID_EQ(first_id, second_id) ? true : false;
+  return UNIQUE_ID_EQ(first_id, second_id);
 }
 
 bool function_id_is_nil(function_id id) {
@@ -288,7 +288,7 @@ struct task_impl {
 };
 
 bool node_ids_equal(node_id first_id, node_id second_id) {
-  return UNIQUE_ID_EQ(first_id, second_id) ? true : false;
+  return UNIQUE_ID_EQ(first_id, second_id);
 }
 
 bool node_id_is_nil(node_id id) {
@@ -320,10 +320,18 @@ scheduling_state task_state(task *task) {
   return task->state;
 }
 
+void task_set_state(task *task, scheduling_state state) {
+  task->state = state;
+}
+
 node_id task_node(task *task) {
   return task->node;
 }
 
+void task_set_node(task *task, node_id node) {
+  task->node = node;
+}
+
 task_spec *task_task_spec(task *task) {
   return &task->spec;
 }
diff --git a/src/common/task.h b/src/common/task.h
index 8d026c05f..ef5aee2bc 100644
--- a/src/common/task.h
+++ b/src/common/task.h
@@ -291,9 +291,15 @@ int64_t task_size(task *task);
 /** The scheduling state of the task. */
 scheduling_state task_state(task *task);
 
+/** Update the schedule state of the task. */
+void task_set_state(task *task, scheduling_state state);
+
 /** Node this task has been assigned to or is running on. */
 node_id task_node(task *task);
 
+/** Set the node for this task. */
+void task_set_node(task *task, node_id node);
+
 /** Task specification of this task. */
 task_spec *task_task_spec(task *task);
 
diff --git a/src/common/test/db_tests.c b/src/common/test/db_tests.c
index f473a9b7a..850cec984 100644
--- a/src/common/test/db_tests.c
+++ b/src/common/test/db_tests.c
@@ -201,25 +201,21 @@ TEST task_table_all_test(void) {
 }
 
 TEST unique_client_id_test(void) {
-  const int num_conns = 50;
+  const int num_conns = 100;
 
+  db_client_id ids[num_conns];
   db_handle *db;
-  pid_t pid = fork();
   for (int i = 0; i < num_conns; ++i) {
     db = db_connect("127.0.0.1", 6379, "plasma_manager", manager_addr,
                     manager_port1);
+    ids[i] = get_db_client_id(db);
     db_disconnect(db);
   }
-  if (pid == 0) {
-    exit(0);
-  } else {
-    wait(NULL);
+  for (int i = 0; i < num_conns; ++i) {
+    for (int j = 0; j < i; ++j) {
+      ASSERT(!db_client_ids_equal(ids[i], ids[j]));
+    }
   }
-
-  db = db_connect("127.0.0.1", 6379, "plasma_manager", manager_addr,
-                  manager_port1);
-  ASSERT_EQ(get_client_id(db), num_conns * 2);
-  db_disconnect(db);
   PASS();
 }
 
diff --git a/src/common/test/redis_tests.c b/src/common/test/redis_tests.c
index 4caec7a53..f51e5457c 100644
--- a/src/common/test/redis_tests.c
+++ b/src/common/test/redis_tests.c
@@ -71,8 +71,7 @@ TEST redis_socket_test(void) {
 void redis_read_callback(event_loop *loop, int fd, void *context, int events) {
   db_handle *db = context;
   char *cmd = read_log_message(fd);
-  redisAsyncCommand(db->context, async_redis_socket_test_callback, NULL, cmd,
-                    db->client_id, 0);
+  redisAsyncCommand(db->context, async_redis_socket_test_callback, NULL, cmd);
   free(cmd);
 }
 
@@ -151,7 +150,7 @@ void logging_read_callback(event_loop *loop,
   db_handle *conn = context;
   char *cmd = read_log_message(fd);
   redisAsyncCommand(conn->context, logging_test_callback, NULL, cmd,
-                    conn->client_id, 0);
+                    (char *) conn->client.id, sizeof(db_client_id));
   free(cmd);
 }
 
diff --git a/src/common/test/test.py b/src/common/test/test.py
index f7d01e41b..b3136de60 100644
--- a/src/common/test/test.py
+++ b/src/common/test/test.py
@@ -1,10 +1,22 @@
 from __future__ import print_function
 
+import numpy as np
 import pickle
 import unittest
 
 import common
 
+ID_SIZE = 20
+
+def random_object_id():
+  return common.ObjectID(np.random.bytes(ID_SIZE))
+
+def random_function_id():
+  return common.ObjectID(np.random.bytes(ID_SIZE))
+
+def random_task_id():
+  return common.ObjectID(np.random.bytes(ID_SIZE))
+
 BASE_SIMPLE_OBJECTS = [
   0, 1, 100000, 0L, 1L, 100000L, 1L << 100, 0.0, 0.5, 0.9, 100000.1, (), [], {},
   "", 990 * "h", u"", 990 * u"h"
@@ -51,10 +63,10 @@ class TestSerialization(unittest.TestCase):
 class TestObjectID(unittest.TestCase):
 
   def test_create_object_id(self):
-    object_id = common.ObjectID(20 * "a")
+    object_id = random_object_id()
 
   def test_cannot_pickle_object_ids(self):
-    object_ids = [common.ObjectID(20 * chr(i)) for i in range(256)]
+    object_ids = [random_object_id() for _ in range(256)]
     def f():
       return object_ids
     def g(val=object_ids):
@@ -71,23 +83,23 @@ class TestObjectID(unittest.TestCase):
     self.assertRaises(Exception, lambda : pickling.dumps(h))
 
   def test_equality_comparisons(self):
-    x1 = common.ObjectID(20 * "a")
-    x2 = common.ObjectID(20 * "a")
-    y1 = common.ObjectID(20 * "b")
-    y2 = common.ObjectID(20 * "b")
+    x1 = common.ObjectID(ID_SIZE * "a")
+    x2 = common.ObjectID(ID_SIZE * "a")
+    y1 = common.ObjectID(ID_SIZE * "b")
+    y2 = common.ObjectID(ID_SIZE * "b")
     self.assertEqual(x1, x2)
     self.assertEqual(y1, y2)
     self.assertNotEqual(x1, y1)
 
-    object_ids1 = [common.ObjectID(20 * chr(i)) for i in range(256)]
-    object_ids2 = [common.ObjectID(20 * chr(i)) for i in range(256)]
+    object_ids1 = [common.ObjectID(ID_SIZE * chr(i)) for i in range(256)]
+    object_ids2 = [common.ObjectID(ID_SIZE * chr(i)) for i in range(256)]
     self.assertEqual(len(set(object_ids1)), 256)
     self.assertEqual(len(set(object_ids1 + object_ids2)), 256)
     self.assertEqual(set(object_ids1), set(object_ids2))
 
   def test_hashability(self):
-    x = common.ObjectID(20 * "a")
-    y = common.ObjectID(20 * "b")
+    x = random_object_id()
+    y = random_object_id()
     {x: y}
     set([x, y])
 
@@ -95,9 +107,9 @@ class TestTask(unittest.TestCase):
 
   def test_create_task(self):
     # TODO(rkn): The function ID should be a FunctionID object, not an ObjectID.
-    parent_id = common.ObjectID(20 * "a")
-    function_id = common.ObjectID(20 * "b")
-    object_ids = [common.ObjectID(20 * chr(i)) for i in range(256)]
+    parent_id = random_task_id()
+    function_id = random_function_id()
+    object_ids = [random_object_id() for _ in range(256)]
     args_list = [
       [],
       1 * [1],
diff --git a/src/global_scheduler/Makefile b/src/global_scheduler/Makefile
new file mode 100644
index 000000000..b98240be2
--- /dev/null
+++ b/src/global_scheduler/Makefile
@@ -0,0 +1,11 @@
+CC = gcc
+CFLAGS = -g -Wall --std=c99 -D_XOPEN_SOURCE=500 -D_POSIX_C_SOURCE=200809L -I. -I../common -I../common/thirdparty -I../common/state
+BUILD = build
+
+all: $(BUILD)/global_scheduler
+
+$(BUILD)/global_scheduler: global_scheduler.c global_scheduler_algorithm.c
+	$(CC) $(CFLAGS) -o $@ global_scheduler.c global_scheduler_algorithm.c ../common/build/libcommon.a ../common/thirdparty/hiredis/libhiredis.a
+
+clean:
+	rm -rf $(BUILD)/*
diff --git a/src/global_scheduler/build/.gitkeep b/src/global_scheduler/build/.gitkeep
new file mode 100644
index 000000000..e69de29bb
diff --git a/src/global_scheduler/global_scheduler.c b/src/global_scheduler/global_scheduler.c
new file mode 100644
index 000000000..e99482161
--- /dev/null
+++ b/src/global_scheduler/global_scheduler.c
@@ -0,0 +1,118 @@
+#include <getopt.h>
+#include <signal.h>
+#include <stdlib.h>
+
+#include "common.h"
+#include "event_loop.h"
+#include "global_scheduler.h"
+#include "global_scheduler_algorithm.h"
+#include "net.h"
+#include "state/db_client_table.h"
+#include "state/table.h"
+#include "state/task_table.h"
+
+/* This is used to define the array of local schedulers used to define the
+ * global_scheduler_state type. */
+UT_icd local_scheduler_icd = {sizeof(local_scheduler), NULL, NULL, NULL};
+
+void assign_task_to_local_scheduler(global_scheduler_state *state,
+                                    task *task,
+                                    node_id node_id) {
+  task_set_state(task, TASK_STATUS_SCHEDULED);
+  task_set_node(task, node_id);
+  retry_info retry = {
+      .num_retries = 0, .timeout = 100, .fail_callback = NULL,
+  };
+  task_table_update(state->db, task, &retry, NULL, NULL);
+}
+
+global_scheduler_state *init_global_scheduler(event_loop *loop,
+                                              const char *redis_addr,
+                                              int redis_port) {
+  global_scheduler_state *state = malloc(sizeof(global_scheduler_state));
+  state->db = db_connect(redis_addr, redis_port, "global_scheduler", "", -1);
+  db_attach(state->db, loop);
+  utarray_new(state->local_schedulers, &local_scheduler_icd);
+  return state;
+}
+
+void free_global_scheduler(global_scheduler_state *state) {
+  db_disconnect(state->db);
+  utarray_free(state->local_schedulers);
+  free(state);
+}
+
+/* We need this code so we can clean up when we get a SIGTERM signal. */
+
+global_scheduler_state *g_state;
+
+void signal_handler(int signal) {
+  if (signal == SIGTERM) {
+    free_global_scheduler(g_state);
+    exit(0);
+  }
+}
+
+/* End of the cleanup code. */
+
+void process_task_waiting(task *task, void *user_context) {
+  global_scheduler_state *state = (global_scheduler_state *) user_context;
+  handle_task_waiting(state, task);
+}
+
+void process_new_db_client(db_client_id db_client_id,
+                           const char *client_type,
+                           void *user_context) {
+  global_scheduler_state *state = (global_scheduler_state *) user_context;
+  if (strcmp(client_type, "photon") == 0) {
+    handle_new_local_scheduler(state, db_client_id);
+  }
+}
+
+void start_server(const char *redis_addr, int redis_port) {
+  event_loop *loop = event_loop_create();
+  g_state = init_global_scheduler(loop, redis_addr, redis_port);
+  /* Generic retry information for notification subscriptions. */
+  retry_info retry = {
+      .num_retries = 0, .timeout = 100, .fail_callback = NULL,
+  };
+  /* TODO(rkn): subscribe to notifications from the object table. */
+  /* Subscribe to notifications about new local schedulers. TODO(rkn): this
+   * needs to also get all of the clients that registered with the database
+   * before this call to subscribe. */
+  db_client_table_subscribe(g_state->db, process_new_db_client,
+                            (void *) g_state, &retry, NULL, NULL);
+  /* Subscribe to notifications about waiting tasks. TODO(rkn): this may need to
+   * get tasks that were submitted to the database before the subscribe. */
+  task_table_subscribe(g_state->db, NIL_ID, TASK_STATUS_WAITING,
+                       process_task_waiting, (void *) g_state, &retry, NULL,
+                       NULL);
+  /* Start the event loop. */
+  event_loop_run(loop);
+}
+
+int main(int argc, char *argv[]) {
+  signal(SIGTERM, signal_handler);
+  /* IP address and port of redis. */
+  char *redis_addr_port = NULL;
+  int c;
+  while ((c = getopt(argc, argv, "s:m:h:p:r:")) != -1) {
+    switch (c) {
+    case 'r':
+      redis_addr_port = optarg;
+      break;
+    default:
+      LOG_ERROR("unknown option %c", c);
+      exit(-1);
+    }
+  }
+  char redis_addr[16];
+  int redis_port;
+  if (!redis_addr_port ||
+      parse_ip_addr_port(redis_addr_port, redis_addr, &redis_port) == -1) {
+    LOG_ERROR(
+        "need to specify redis address like 127.0.0.1:6379 with -r switch");
+    exit(-1);
+  }
+  start_server(redis_addr, redis_port);
+}
diff --git a/src/global_scheduler/global_scheduler.h b/src/global_scheduler/global_scheduler.h
new file mode 100644
index 000000000..d9b97ea67
--- /dev/null
+++ b/src/global_scheduler/global_scheduler.h
@@ -0,0 +1,28 @@
+#ifndef GLOBAL_SCHEDULER_H
+#define GLOBAL_SCHEDULER_H
+
+#include "task.h"
+
+#include "state/db.h"
+#include "utarray.h"
+
+/** Contains all information that is associated with a local scheduler. */
+typedef struct {
+  /** The ID of the local scheduler in Redis. */
+  db_client_id id;
+} local_scheduler;
+
+typedef struct {
+  /** The global scheduler event loop. */
+  event_loop *loop;
+  /** The global state store database. */
+  db_handle *db;
+  /** The local schedulers that are connected to Redis. */
+  UT_array *local_schedulers;
+} global_scheduler_state;
+
+void assign_task_to_local_scheduler(global_scheduler_state *state,
+                                    task *task,
+                                    node_id node_id);
+
+#endif /* GLOBAL_SCHEDULER_H */
diff --git a/src/global_scheduler/global_scheduler_algorithm.c b/src/global_scheduler/global_scheduler_algorithm.c
new file mode 100644
index 000000000..7c2415068
--- /dev/null
+++ b/src/global_scheduler/global_scheduler_algorithm.c
@@ -0,0 +1,29 @@
+#include "task.h"
+#include "task_table.h"
+
+#include "global_scheduler_algorithm.h"
+
+void handle_task_waiting(global_scheduler_state *state, task *task) {
+  if (utarray_len(state->local_schedulers) > 0) {
+    local_scheduler *scheduler =
+        (local_scheduler *) utarray_eltptr(state->local_schedulers, 0);
+    assign_task_to_local_scheduler(state, task, scheduler->id);
+  } else {
+    CHECKM(0, "We currently don't handle this case.");
+  }
+}
+
+void handle_object_available(global_scheduler_state *state,
+                             object_id object_id) {
+  /* Do nothing for now. */
+}
+
+void handle_local_scheduler_heartbeat(global_scheduler_state *state) {
+  /* Do nothing for now. */
+}
+
+void handle_new_local_scheduler(global_scheduler_state *state,
+                                db_client_id db_client_id) {
+  local_scheduler local_scheduler = {.id = db_client_id};
+  utarray_push_back(state->local_schedulers, &local_scheduler);
+}
diff --git a/src/global_scheduler/global_scheduler_algorithm.h b/src/global_scheduler/global_scheduler_algorithm.h
new file mode 100644
index 000000000..86f089006
--- /dev/null
+++ b/src/global_scheduler/global_scheduler_algorithm.h
@@ -0,0 +1,57 @@
+#ifndef GLOBAL_SCHEDULER_ALGORITHM_H
+#define GLOBAL_SCHEDULER_ALGORITHM_H
+
+#include "common.h"
+#include "global_scheduler.h"
+#include "task.h"
+
+/* ==== The scheduling algorithm ====
+ *
+ * This file contains declaration for all functions and data structures that
+ * need to be provided if you want to implement a new algorithm for the global
+ * scheduler.
+ *
+ */
+
+/**
+ * Assign the task to a local scheduler. At the moment, this simply assigns the
+ * task to the first local scheduler and if there are no local schedulers it
+ * fails.
+ *
+ * @param state The global scheduler state.
+ * @param task The task that is waiting to be scheduled.
+ * @return Void.
+ */
+void handle_task_waiting(global_scheduler_state *state, task *task);
+
+/**
+ * Handle the fact that a new object is available.
+ *
+ * @param state The global scheduler state.
+ * @param object_id The ID of the object that is now available.
+ * @return Void.
+ */
+void handle_object_available(global_scheduler_state *state,
+                             object_id object_id);
+
+/**
+ * Handle a heartbeat message from a local scheduler. TODO(rkn): this is a
+ * placeholder for now.
+ *
+ * @param state The global scheduler state.
+ * @return Void.
+ */
+void handle_local_scheduler_heartbeat(global_scheduler_state *state);
+
+/**
+ * Handle the presence of a new local scheduler. Currently, this just adds the
+ * local scheduler to a queue of local schedulers.
+ *
+ * @param state The global scheduler state.
+ * @param The db client ID of the new local scheduler.
+ * @return Void.
+ */
+void handle_new_local_scheduler(global_scheduler_state *state,
+                                db_client_id db_client_id);
+
+#endif /* GLOBAL_SCHEDULER_ALGORITHM_H */
diff --git a/src/global_scheduler/lib/python/global_scheduler_services.py b/src/global_scheduler/lib/python/global_scheduler_services.py
new file mode 100644
index 000000000..5191ffada
--- /dev/null
+++ b/src/global_scheduler/lib/python/global_scheduler_services.py
@@ -0,0 +1,33 @@
+from __future__ import print_function
+
+import os
+import subprocess
+import time
+
+def start_global_scheduler(redis_address, use_valgrind=False, use_profiler=False):
+  """Start a global scheduler process.
+
+  Args:
+    redis_address (str): The address of the Redis instance.
+    use_valgrind (bool): True if the global scheduler should be started inside
+      of valgrind. If this is True, use_profiler must be False.
+    use_profiler (bool): True if the global scheduler should be started inside a
+      profiler. If this is True, use_valgrind must be False.
+
+  Return:
+    The process ID of the global scheduler process.
+  """
+  if use_valgrind and use_profiler:
+    raise Exception("Cannot use valgrind and profiler at the same time.")
+  global_scheduler_executable = os.path.join(os.path.abspath(os.path.dirname(__file__)), "../../build/global_scheduler")
+  command = [global_scheduler_executable, "-r", redis_address]
+  if use_valgrind:
+    pid = subprocess.Popen(["valgrind", "--track-origins=yes", "--leak-check=full", "--show-leak-kinds=all", "--error-exitcode=1"] + command)
+    time.sleep(1.0)
+  elif use_profiler:
+    pid = subprocess.Popen(["valgrind", "--tool=callgrind"] + command)
+    time.sleep(1.0)
+  else:
+    pid = subprocess.Popen(command)
+    time.sleep(0.1)
+  return pid
diff --git a/src/global_scheduler/test/test.py b/src/global_scheduler/test/test.py
new file mode 100644
index 000000000..14d47a86b
--- /dev/null
+++ b/src/global_scheduler/test/test.py
@@ -0,0 +1,123 @@
+from __future__ import print_function
+
+import numpy as np
+import os
+import random
+import redis
+import signal
+import subprocess
+import sys
+import threading
+import time
+import unittest
+
+import global_scheduler
+import photon
+import plasma
+
+USE_VALGRIND = False
+PLASMA_STORE_MEMORY = 1000000000
+ID_SIZE = 20
+
+# These constants must match the schedulign state enum in task.h.
+TASK_STATUS_WAITING = 1
+TASK_STATUS_SCHEDULED = 2
+TASK_STATUS_RUNNING = 4
+TASK_STATUS_DONE = 8
+
+def random_object_id():
+  return photon.ObjectID(np.random.bytes(ID_SIZE))
+
+def random_task_id():
+  return photon.ObjectID(np.random.bytes(ID_SIZE))
+
+def random_function_id():
+  return photon.ObjectID(np.random.bytes(ID_SIZE))
+
+def new_port():
+  return random.randint(10000, 65535)
+
+class TestGlobalScheduler(unittest.TestCase):
+
+  def setUp(self):
+    # Start a Redis server.
+    redis_path = os.path.join(os.path.abspath(os.path.dirname(__file__)), "../../common/thirdparty/redis-3.2.3/src/redis-server")
+    node_ip_address = "127.0.0.1"
+    redis_port = new_port()
+    redis_address = "{}:{}".format(node_ip_address, redis_port)
+    self.redis_process = subprocess.Popen([redis_path, "--port", str(redis_port), "--loglevel", "warning"])
+    time.sleep(0.1)
+    # Create a Redis client.
+    self.redis_client = redis.StrictRedis(host=node_ip_address, port=redis_port)
+    # Start the global scheduler.
+    self.p1 = global_scheduler.start_global_scheduler(redis_address, USE_VALGRIND)
+    # Start the Plasma store.
+    plasma_store_name, self.p2 = plasma.start_plasma_store()
+    # Start the local scheduler.
+    local_scheduler_name, self.p3 = photon.start_local_scheduler(plasma_store_name, redis_address=redis_address)
+    # Connect to the scheduler.
+    self.photon_client = photon.PhotonClient(local_scheduler_name)
+
+  def tearDown(self):
+    # Kill the global scheduler.
+    if USE_VALGRIND:
+      self.p1.send_signal(signal.SIGTERM)
+      self.p1.wait()
+      os._exit(self.p1.returncode)
+    else:
+      self.p1.kill()
+    self.p2.kill()
+    self.p3.kill()
+    # Kill Redis. In the event that we are using valgrind, this needs to happen
+    # after we kill the global scheduler.
+    self.redis_process.kill()
+
+  def test_redis_contents(self):
+    # There should be two db clients, the global scheduler and the local
+    # scheduler.
+    self.assertEqual(len(self.redis_client.keys("db_clients*")), 2)
+    # There should not be anything else in Redis yet.
+    self.assertEqual(len(self.redis_client.keys("*")), 2)
+
+    # Submit a task to Redis.
+    task = photon.Task(random_function_id(), [], 0, random_task_id(), 0)
+    self.photon_client.submit(task)
+    # There should now be a task in Redis, and it should get assigned to the
+    # local scheduler
+    while True:
+      task_entries = self.redis_client.keys("task*")
+      self.assertLessEqual(len(task_entries), 1)
+      if len(task_entries) == 1:
+        task_contents = self.redis_client.hgetall(task_entries[0])
+        task_status = int(task_contents["state"])
+        self.assertTrue(task_status in [TASK_STATUS_WAITING, TASK_STATUS_SCHEDULED])
+        if task_status == TASK_STATUS_SCHEDULED:
+          break
+      print("The task has not been scheduled yet, trying again.")
+
+    # Submit a bunch of tasks to Redis.
+    num_tasks = 1000
+    for _ in range(num_tasks):
+      task = photon.Task(random_function_id(), [], 0, random_task_id(), 0)
+      self.photon_client.submit(task)
+    # Check that there are the correct number of tasks in Redis and that they
+    # all get assigned to the local scheduler.
+    while True:
+      task_entries = self.redis_client.keys("task*")
+      self.assertLessEqual(len(task_entries), num_tasks + 1)
+      if len(task_entries) == num_tasks + 1:
+        task_contents = [self.redis_client.hgetall(task_entries[i]) for i in range(len(task_entries))]
+        task_statuses = [int(contents["state"]) for contents in task_contents]
+        self.assertTrue(all([status in [TASK_STATUS_WAITING, TASK_STATUS_SCHEDULED] for status in task_statuses]))
+        if all([status == TASK_STATUS_SCHEDULED for status in task_statuses]):
+          break
+      print("The tasks have not been scheduled yet, trying again.")
+
+if __name__ == "__main__":
+  if len(sys.argv) > 1:
+    # pop the argument so we don't mess with unittest's own argument parser
+    arg = sys.argv.pop()
+    if arg == "valgrind":
+      USE_VALGRIND = True
+      print("Using valgrind for tests")
+  unittest.main(verbosity=2)
diff --git a/src/photon/Makefile b/src/photon/Makefile
index 31fdee210..9d7db8eb7 100644
--- a/src/photon/Makefile
+++ b/src/photon/Makefile
@@ -14,7 +14,7 @@ common: FORCE
 	cd ../common; make
 
 clean:
-	rm -r $(BUILD)/*
-	rm *.o
+	rm -rf $(BUILD)/*
+	rm -f *.o
 
 FORCE:
diff --git a/src/photon/photon/__init__.py b/src/photon/photon/__init__.py
index 27f7c578f..d34d8d3e1 100644
--- a/src/photon/photon/__init__.py
+++ b/src/photon/photon/__init__.py
@@ -1 +1,2 @@
 from libphoton import *
+from photon_services import *
diff --git a/src/photon/photon/photon_services.py b/src/photon/photon/photon_services.py
new file mode 100644
index 000000000..bc7e8f362
--- /dev/null
+++ b/src/photon/photon/photon_services.py
@@ -0,0 +1,43 @@
+from __future__ import print_function
+
+import os
+import random
+import subprocess
+import time
+
+def random_name():
+  return str(random.randint(0, 99999999))
+
+def start_local_scheduler(plasma_store_name, redis_address=None, use_valgrind=False, use_profiler=False):
+  """Start a local scheduler process.
+
+  Args:
+    plasma_store_name (str): The name of the plasma store socket to connect to.
+    redis_address (str): The address of the Redis instance to connect to. If
+      this is not provided, then the local scheduler will not connect to Redis.
+    use_valgrind (bool): True if the local scheduler should be started inside of
+      valgrind. If this is True, use_profiler must be False.
+    use_profiler (bool): True if the local scheduler should be started inside a
+      profiler. If this is True, use_valgrind must be False.
+
+  Return:
+    A tuple of the name of the local scheduler socket and the process ID of the
+      local scheduler process.
+  """
+  if use_valgrind and use_profiler:
+    raise Exception("Cannot use valgrind and profiler at the same time.")
+  local_scheduler_executable = os.path.join(os.path.dirname(os.path.abspath(__file__)), "../build/photon_scheduler")
+  local_scheduler_name = "/tmp/scheduler{}".format(random_name())
+  command = [local_scheduler_executable, "-s", local_scheduler_name, "-p", plasma_store_name]
+  if redis_address is not None:
+    command += ["-r", redis_address]
+  if use_valgrind:
+    pid = subprocess.Popen(["valgrind", "--track-origins=yes", "--leak-check=full", "--show-leak-kinds=all", "--error-exitcode=1"] + command)
+    time.sleep(1.0)
+  elif use_profiler:
+    pid = subprocess.Popen(["valgrind", "--tool=callgrind"] + command)
+    time.sleep(1.0)
+  else:
+    pid = subprocess.Popen(command)
+    time.sleep(0.1)
+  return local_scheduler_name, pid
diff --git a/src/photon/photon_algorithm.c b/src/photon/photon_algorithm.c
index ab26e9c8e..1da91a13b 100644
--- a/src/photon/photon_algorithm.c
+++ b/src/photon/photon_algorithm.c
@@ -16,15 +16,19 @@ const retry_info photon_retry = {
 };
 
 typedef struct task_queue_entry {
-  task *task;
+  /** The task that is queued. */
+  task_spec *spec;
+  /** True if this task was assigned to this local scheduler by the global
+   *  scheduler and false otherwise. */
+  bool from_global_scheduler;
   struct task_queue_entry *prev;
   struct task_queue_entry *next;
 } task_queue_entry;
 
 typedef struct {
-  /* Object id of this object. */
+  /** Object id of this object. */
   object_id object_id;
-  /* Handle for the uthash table. */
+  /** Handle for the uthash table. */
   UT_hash_handle handle;
 } available_object;
 
@@ -54,7 +58,7 @@ void free_scheduler_state(scheduler_state *s) {
   task_queue_entry *elt, *tmp1;
   DL_FOREACH_SAFE(s->task_queue, elt, tmp1) {
     DL_DELETE(s->task_queue, elt);
-    free(elt->task);
+    free_task_spec(elt->spec);
     free(elt);
   }
   utarray_free(s->available_workers);
@@ -101,67 +105,99 @@ bool can_run(scheduler_state *s, task_spec *task) {
  * @return This returns 1 if it successfully assigned a task to the worker,
  *         otherwise it returns 0.
  */
-int find_and_schedule_task_if_possible(scheduler_info *info,
-                                       scheduler_state *state,
-                                       int worker_index) {
+bool find_and_schedule_task_if_possible(scheduler_info *info,
+                                        scheduler_state *state,
+                                        int worker_index) {
   task_queue_entry *elt, *tmp;
-  task_spec *spec;
-  int found_task_to_schedule = 0;
+  bool found_task_to_schedule = false;
   /* Find the first task whose dependencies are available locally. */
   DL_FOREACH_SAFE(state->task_queue, elt, tmp) {
-    spec = task_task_spec(elt->task);
-    if (can_run(state, spec)) {
-      found_task_to_schedule = 1;
+    if (can_run(state, elt->spec)) {
+      found_task_to_schedule = true;
       break;
     }
   }
   if (found_task_to_schedule) {
     /* This task's dependencies are available locally, so assign the task to the
      * worker. */
-    assign_task_to_worker(info, spec, worker_index);
+    assign_task_to_worker(info, elt->spec, worker_index,
+                          elt->from_global_scheduler);
     /* Update the task queue data structure and free the task. */
     DL_DELETE(state->task_queue, elt);
-    free(elt->task);
+    free_task_spec(elt->spec);
     free(elt);
   }
   return found_task_to_schedule;
 }
 
+void run_task_immediately(scheduler_info *info,
+                          scheduler_state *s,
+                          task_spec *spec,
+                          bool from_global_scheduler) {
+  /* Get the last available worker in the available worker queue. */
+  int *worker_index = (int *) utarray_back(s->available_workers);
+  /* Tell the available worker to execute the task. */
+  assign_task_to_worker(info, spec, *worker_index, from_global_scheduler);
+  /* Remove the available worker from the queue and free the struct. */
+  utarray_pop_back(s->available_workers);
+}
+
+void queue_task_locally(scheduler_info *info,
+                        scheduler_state *s,
+                        task_spec *spec,
+                        bool from_global_scheduler) {
+  /* Copy the spec and add it to the task queue. The allocated spec will be
+   * freed when it is assigned to a worker. */
+  task_queue_entry *elt = malloc(sizeof(task_queue_entry));
+  elt->spec = malloc(task_spec_size(spec));
+  memcpy(elt->spec, spec, task_spec_size(spec));
+  elt->from_global_scheduler = from_global_scheduler;
+  DL_APPEND(s->task_queue, elt);
+}
+
+void give_task_to_global_scheduler(scheduler_info *info,
+                                   scheduler_state *s,
+                                   task_spec *spec,
+                                   bool from_global_scheduler) {
+  /* Pass on the task to the global scheduler. */
+  DCHECK(!from_global_scheduler);
+  task *task = alloc_task(spec, TASK_STATUS_WAITING, NIL_ID);
+  DCHECK(info->db != NULL);
+  task_table_add_task(info->db, task, (retry_info *) &photon_retry, NULL, NULL);
+  free_task(task);
+}
+
 void handle_task_submitted(scheduler_info *info,
                            scheduler_state *s,
                            task_spec *spec) {
-  /* Create a unique task instance ID. This is different from the task ID and
-   * is used to distinguish between potentially multiple executions of the
-   * task. */
-  task *task = alloc_task(spec, TASK_STATUS_WAITING, NIL_ID);
   /* If this task's dependencies are available locally, and if there is an
-   * available worker, then assign this task to an available worker. Otherwise,
-   * add this task to the local task queue. */
-  int schedule_locally =
-      (utarray_len(s->available_workers) > 0) && can_run(s, spec);
-  if (schedule_locally) {
-    /* Get the last available worker in the available worker queue. */
-    int *worker_index = (int *) utarray_back(s->available_workers);
-    /* Tell the available worker to execute the task. */
-    assign_task_to_worker(info, spec, *worker_index);
-    /* Remove the available worker from the queue and free the struct. */
-    utarray_pop_back(s->available_workers);
+   * available worker, then assign this task to an available worker. If we
+   * cannot assign the task to a worker immediately, we either queue the task in
+   * the local task queue or we pass the task to the global scheduler. For now,
+   * we pass the task along to the global scheduler if there is one. */
+  if ((utarray_len(s->available_workers) > 0) && can_run(s, spec)) {
+    run_task_immediately(info, s, spec, false);
+  } else if (info->db == NULL) {
+    queue_task_locally(info, s, spec, false);
   } else {
-    /* Add the task to the task queue. This passes ownership of the task queue.
-     * And the task will be freed when it is assigned to a worker. */
-    task_queue_entry *elt = malloc(sizeof(task_queue_entry));
-    elt->task = task;
-    DL_APPEND(s->task_queue, elt);
+    give_task_to_global_scheduler(info, s, spec, false);
   }
-  /* Submit the task to redis. */
-  /* TODO(swang): This should be task_table_update if the task is already in the
-   * log. */
-  task_table_add_task(info->db, task, (retry_info *) &photon_retry, NULL, NULL);
-  if (schedule_locally) {
-    /* If the task was scheduled locally, we need to free it. Otherwise,
-     * ownership of the task is passed to the task_queue, and it will be freed
-     * when it is assigned to a worker. */
-    free_task(task);
+}
+
+void handle_task_scheduled(scheduler_info *info,
+                           scheduler_state *s,
+                           task_spec *spec) {
+  /* This callback handles tasks that were assigned to this local scheduler by
+   * the global scheduler, so we can safely assert that there is a connection
+   * to the database. */
+  DCHECK(info->db != NULL);
+  /* If this task's dependencies are available locally, and if there is an
+   * available worker, then assign this task to an available worker. If we
+   * cannot assign the task to a worker immediately, queue the task locally. */
+  if ((utarray_len(s->available_workers) > 0) && can_run(s, spec)) {
+    run_task_immediately(info, s, spec, true);
+  } else {
+    queue_task_locally(info, s, spec, true);
   }
 }
 
diff --git a/src/photon/photon_algorithm.h b/src/photon/photon_algorithm.h
index ff2f023e4..cff36b27d 100644
--- a/src/photon/photon_algorithm.h
+++ b/src/photon/photon_algorithm.h
@@ -54,9 +54,9 @@ void handle_task_submitted(scheduler_info *info,
  * @param task Task that is assigned by the global scheduler.
  * @return Void.
  */
-void handle_task_assigned(scheduler_info *info,
-                          scheduler_state *state,
-                          task_spec *task);
+void handle_task_scheduled(scheduler_info *info,
+                           scheduler_state *state,
+                           task_spec *spec);
 
 /**
  * This function is called if a new object becomes available in the local
diff --git a/src/photon/photon_scheduler.c b/src/photon/photon_scheduler.c
index 25c54e8a3..7fcff070f 100644
--- a/src/photon/photon_scheduler.c
+++ b/src/photon/photon_scheduler.c
@@ -69,10 +69,14 @@ local_scheduler_state *init_local_scheduler(event_loop *loop,
   /* Add scheduler info. */
   state->scheduler_info = malloc(sizeof(scheduler_info));
   utarray_new(state->scheduler_info->workers, &worker_icd);
-  /* Connect to Redis. */
-  state->scheduler_info->db =
-      db_connect(redis_addr, redis_port, "photon", "", -1);
-  db_attach(state->scheduler_info->db, loop);
+  /* Connect to Redis if a Redis address is provided. */
+  if (redis_addr != NULL) {
+    state->scheduler_info->db =
+        db_connect(redis_addr, redis_port, "photon", "", -1);
+    db_attach(state->scheduler_info->db, loop);
+  } else {
+    state->scheduler_info->db = NULL;
+  }
   /* Add scheduler state. */
   state->scheduler_state = make_scheduler_state();
   utarray_new(state->input_buffer, &byte_icd);
@@ -80,7 +84,9 @@ local_scheduler_state *init_local_scheduler(event_loop *loop,
 };
 
 void free_local_scheduler(local_scheduler_state *s) {
-  db_disconnect(s->scheduler_info->db);
+  if (s->scheduler_info->db != NULL) {
+    db_disconnect(s->scheduler_info->db);
+  }
   plasma_disconnect(s->plasma_conn);
   worker_index *current_worker_index, *temp_worker_index;
   HASH_ITER(hh, s->worker_index, current_worker_index, temp_worker_index) {
@@ -97,10 +103,25 @@ void free_local_scheduler(local_scheduler_state *s) {
 
 void assign_task_to_worker(scheduler_info *info,
                            task_spec *spec,
-                           int worker_index) {
+                           int worker_index,
+                           bool from_global_scheduler) {
   CHECK(worker_index < utarray_len(info->workers));
   worker *w = (worker *) utarray_eltptr(info->workers, worker_index);
   write_message(w->sock, EXECUTE_TASK, task_spec_size(spec), (uint8_t *) spec);
+  /* Update the global task table. */
+  if (info->db != NULL) {
+    retry_info retry = {
+        .num_retries = 0, .timeout = 100, .fail_callback = NULL,
+    };
+    task *task =
+        alloc_task(spec, TASK_STATUS_RUNNING, get_db_client_id(info->db));
+    if (from_global_scheduler) {
+      task_table_update(info->db, task, (retry_info *) &retry, NULL, NULL);
+    } else {
+      task_table_add_task(info->db, task, (retry_info *) &retry, NULL, NULL);
+    }
+    free_task(task);
+  }
 }
 
 void process_plasma_notification(event_loop *loop,
@@ -177,6 +198,11 @@ void signal_handler(int signal) {
 
 /* End of the cleanup code. */
 
+void handle_task_scheduled_callback(task *original_task, void *user_context) {
+  handle_task_scheduled(g_state->scheduler_info, g_state->scheduler_state,
+                        task_task_spec(original_task));
+}
+
 void start_server(const char *socket_name,
                   const char *redis_addr,
                   int redis_port,
@@ -186,9 +212,23 @@ void start_server(const char *socket_name,
   g_state =
       init_local_scheduler(loop, redis_addr, redis_port, plasma_socket_name);
 
-  /* Run event loop. */
+  /* Register a callback for registering new clients. */
   event_loop_add_file(loop, fd, EVENT_LOOP_READ, new_client_connection,
                       g_state);
+  /* Subscribe to receive notifications about tasks that are assigned to this
+   * local scheduler by the global scheduler. TODO(rkn): we also need to get any
+   * tasks that were assigned to this local scheduler before the call to
+   * subscribe. */
+  retry_info retry = {
+      .num_retries = 0, .timeout = 100, .fail_callback = NULL,
+  };
+  if (g_state->scheduler_info->db != NULL) {
+    task_table_subscribe(g_state->scheduler_info->db,
+                         get_db_client_id(g_state->scheduler_info->db),
+                         TASK_STATUS_SCHEDULED, handle_task_scheduled_callback,
+                         NULL, &retry, NULL, NULL);
+  }
+  /* Run event loop. */
   event_loop_run(loop);
 }
 
@@ -222,15 +262,21 @@ int main(int argc, char *argv[]) {
   if (!plasma_socket_name) {
     LOG_FATAL("please specify socket for connecting to Plasma with -p switch");
   }
-  /* Parse the Redis address into an IP address and a port. */
-  char redis_addr[16] = {0};
-  char redis_port[6] = {0};
-  if (!redis_addr_port ||
-      sscanf(redis_addr_port, "%15[0-9.]:%5[0-9]", redis_addr, redis_port) !=
-          2) {
-    LOG_FATAL(
-        "need to specify redis address like 127.0.0.1:6379 with -r switch");
+  if (!redis_addr_port) {
+    /* Start the local scheduler without connecting to Redis. In this case, all
+     * submitted tasks will be queued and scheduled locally. */
+    start_server(scheduler_socket_name, NULL, -1, plasma_socket_name);
+  } else {
+    /* Parse the Redis address into an IP address and a port. */
+    char redis_addr[16] = {0};
+    char redis_port[6] = {0};
+    if (sscanf(redis_addr_port, "%15[0-9.]:%5[0-9]", redis_addr, redis_port) !=
+        2) {
+      LOG_FATAL(
+          "if a redis address is provided with the -r switch, it should be "
+          "formatted like 127.0.0.1:6379");
+    }
+    start_server(scheduler_socket_name, &redis_addr[0], atoi(redis_port),
+                 plasma_socket_name);
   }
-  start_server(scheduler_socket_name, &redis_addr[0], atoi(redis_port),
-               plasma_socket_name);
 }
diff --git a/src/photon/photon_scheduler.h b/src/photon/photon_scheduler.h
index 3ebb03432..0c81ce97e 100644
--- a/src/photon/photon_scheduler.h
+++ b/src/photon/photon_scheduler.h
@@ -28,11 +28,14 @@ void new_client_connection(event_loop *loop,
  * @param info
  * @param task The task that is submitted to the worker.
  * @param worker_index The index of the worker the task is submitted to.
+ * @param from_global_scheduler True if the task was assigned to the local
+ *        scheduler by the global scheduler and false otherwise.
  * @return Void.
  */
 void assign_task_to_worker(scheduler_info *info,
                            task_spec *task,
-                           int worker_index);
+                           int worker_index,
+                           bool from_global_scheduler);
 
 /**
  * This is the callback that is used to process a notification from the Plasma
diff --git a/src/photon/setup.py b/src/photon/setup.py
deleted file mode 100644
index 863649863..000000000
--- a/src/photon/setup.py
+++ /dev/null
@@ -1,23 +0,0 @@
-import subprocess
-
-from setuptools import setup, find_packages
-import setuptools.command.install as _install
-
-class install(_install.install):
-  def run(self):
-    subprocess.check_call(["make"])
-    subprocess.check_call(["cmake", ".."], cwd="build")
-    subprocess.check_call(["make", "install"], cwd="build")
-    # Calling _install.install.run(self) does not fetch required packages and
-    # instead performs an old-style install. See command/install.py in
-    # setuptools. So, calling do_egg_install() manually here.
-    self.do_egg_install()
-
-setup(name="photon",
-      version="0.1",
-      description="Photon library for Ray",
-      packages=find_packages(),
-      package_data={"photon": ["libphoton.so"]},
-      cmdclass={"install": install},
-      include_package_data=True,
-      zip_safe=False)
diff --git a/src/photon/test/test.py b/src/photon/test/test.py
index 4a8dc2c1d..427d69f50 100644
--- a/src/photon/test/test.py
+++ b/src/photon/test/test.py
@@ -1,66 +1,55 @@
 from __future__ import print_function
 
+import numpy as np
 import os
+import random
 import signal
 import subprocess
 import sys
-import unittest
-import random
 import threading
 import time
+import unittest
 
 import photon
 import plasma
 
 USE_VALGRIND = False
-PLASMA_STORE_MEMORY = 1000000000
+ID_SIZE = 20
 
 def random_object_id():
-  return photon.ObjectID("".join([chr(random.randint(0, 255)) for _ in range(plasma.PLASMA_ID_SIZE)]))
+  return photon.ObjectID(np.random.bytes(ID_SIZE))
+
+def random_task_id():
+  return photon.ObjectID(np.random.bytes(ID_SIZE))
+
+def random_function_id():
+  return photon.ObjectID(np.random.bytes(ID_SIZE))
 
 class TestPhotonClient(unittest.TestCase):
 
   def setUp(self):
-    # Start Redis.
-    redis_executable = os.path.join(os.path.abspath(os.path.dirname(__file__)), "../../common/thirdparty/redis-3.2.3/src/redis-server")
-    self.p1 = subprocess.Popen([redis_executable, "--loglevel", "warning"])
-    # Start Plasma.
-    plasma_executable = os.path.join(os.path.abspath(os.path.dirname(__file__)), "../../plasma/build/plasma_store")
-    plasma_socket = "/tmp/plasma_store{}".format(random.randint(0, 10000))
-    self.p2 = subprocess.Popen([plasma_executable, "-s", plasma_socket, "-m", str(PLASMA_STORE_MEMORY)])
-    time.sleep(0.1)
-    self.plasma_client = plasma.PlasmaClient(plasma_socket)
-    scheduler_executable = os.path.join(os.path.abspath(os.path.dirname(__file__)), "../build/photon_scheduler")
-    scheduler_name = "/tmp/scheduler{}".format(random.randint(0, 10000))
-    command = [scheduler_executable, "-s", scheduler_name, "-r", "127.0.0.1:6379", "-p", plasma_socket]
-    if USE_VALGRIND:
-      self.p3 = subprocess.Popen(["valgrind", "--track-origins=yes", "--leak-check=full", "--show-leak-kinds=all", "--error-exitcode=1"] + command)
-    else:
-      self.p3 = subprocess.Popen(command)
-    if USE_VALGRIND:
-      time.sleep(1.0)
-    else:
-      time.sleep(0.1)
+    # Start Plasma store.
+    plasma_store_name, self.p1 = plasma.start_plasma_store()
+    self.plasma_client = plasma.PlasmaClient(plasma_store_name)
+    # Start a local scheduler.
+    scheduler_name, self.p2 = photon.start_local_scheduler(plasma_store_name, use_valgrind=USE_VALGRIND)
     # Connect to the scheduler.
     self.photon_client = photon.PhotonClient(scheduler_name)
 
   def tearDown(self):
-    # Kill the Redis server.
-    self.p1.kill()
     # Kill Plasma.
-    self.p2.kill()
+    self.p1.kill()
     # Kill the local scheduler.
     if USE_VALGRIND:
-      self.p3.send_signal(signal.SIGTERM)
-      self.p3.wait()
-      os._exit(self.p3.returncode)
+      self.p2.send_signal(signal.SIGTERM)
+      self.p2.wait()
+      os._exit(self.p2.returncode)
     else:
-      self.p3.kill()
+      self.p2.kill()
 
   def test_submit_and_get_task(self):
-    # TODO(rkn): This should be a FunctionID.
-    function_id = photon.ObjectID(20 * "a")
-    object_ids = [photon.ObjectID(20 * chr(i)) for i in range(256)]
+    function_id = random_function_id()
+    object_ids = [random_object_id() for i in range(256)]
     # Create and seal the objects in the object store so that we can schedule
     # all of the subsequent tasks.
     for object_id in object_ids:
@@ -98,7 +87,7 @@ class TestPhotonClient(unittest.TestCase):
 
     for args in args_list:
       for num_return_vals in [0, 1, 2, 3, 5, 10, 100]:
-        task = photon.Task(function_id, args, num_return_vals, random_object_id(), 0)
+        task = photon.Task(function_id, args, num_return_vals, random_task_id(), 0)
         # Submit a task.
         self.photon_client.submit(task)
         # Get the task.
@@ -117,7 +106,7 @@ class TestPhotonClient(unittest.TestCase):
     # Submit all of the tasks.
     for args in args_list:
       for num_return_vals in [0, 1, 2, 3, 5, 10, 100]:
-        task = photon.Task(function_id, args, num_return_vals, random_object_id(), 0)
+        task = photon.Task(function_id, args, num_return_vals, random_task_id(), 0)
         self.photon_client.submit(task)
     # Get all of the tasks.
     for args in args_list:
@@ -126,10 +115,8 @@ class TestPhotonClient(unittest.TestCase):
 
   def test_scheduling_when_objects_ready(self):
     # Create a task and submit it.
-    object_id = photon.ObjectID(20 * chr(0))
-    # TODO(rkn): This should be a FunctionID.
-    function_id = photon.ObjectID(20 * "a")
-    task = photon.Task(function_id, [object_id], 0, random_object_id(), 0)
+    object_id = random_object_id()
+    task = photon.Task(random_function_id(), [object_id], 0, random_task_id(), 0)
     self.photon_client.submit(task)
     # Launch a thread to get the task.
     def get_task():
diff --git a/src/plasma/Makefile b/src/plasma/Makefile
index 38ffde588..1e33f376c 100644
--- a/src/plasma/Makefile
+++ b/src/plasma/Makefile
@@ -11,7 +11,7 @@ debug: all
 
 clean:
 	cd ../common; make clean
-	rm -r $(BUILD)/*
+	rm -rf $(BUILD)/*
 
 $(BUILD)/manager_tests: test/manager_tests.c plasma.h plasma_client.h plasma_client.c plasma_manager.h plasma_manager.c fling.h fling.c common
 	$(CC) $(CFLAGS) $(TEST_CFLAGS) -o $@ test/manager_tests.c plasma_manager.c plasma_client.c fling.c ../common/build/libcommon.a ../common/thirdparty/hiredis/libhiredis.a
diff --git a/src/plasma/lib/python/plasma.py b/src/plasma/lib/python/plasma.py
index 3390013e5..e35225e8f 100644
--- a/src/plasma/lib/python/plasma.py
+++ b/src/plasma/lib/python/plasma.py
@@ -218,24 +218,58 @@ class PlasmaClient(object):
         break
     return message_data
 
-def start_plasma_manager(store_name, manager_name, redis_address, num_retries=20, use_valgrind=False, run_profiler=False):
+DEFAULT_PLASMA_STORE_MEMORY = 10 ** 9
+
+def random_name():
+  return str(random.randint(0, 99999999))
+
+def start_plasma_store(plasma_store_memory=DEFAULT_PLASMA_STORE_MEMORY, use_valgrind=False, use_profiler=False):
+  """Start a plasma store process.
+
+  Args:
+    use_valgrind (bool): True if the plasma store should be started inside of
+      valgrind. If this is True, use_profiler must be False.
+    use_profiler (bool): True if the plasma store should be started inside a
+      profiler. If this is True, use_valgrind must be False.
+
+  Return:
+    A tuple of the name of the plasma store socket and the process ID of the
+      plasma store process.
+  """
+  if use_valgrind and use_profiler:
+    raise Exception("Cannot use valgrind and profiler at the same time.")
+  plasma_store_executable = os.path.join(os.path.abspath(os.path.dirname(__file__)), "../../build/plasma_store")
+  plasma_store_name = "/tmp/scheduler{}".format(random_name())
+  command = [plasma_store_executable, "-s", plasma_store_name, "-m", str(plasma_store_memory)]
+  if use_valgrind:
+    pid = subprocess.Popen(["valgrind", "--track-origins=yes", "--leak-check=full", "--show-leak-kinds=all", "--error-exitcode=1"] + command)
+    time.sleep(1.0)
+  elif use_profiler:
+    pid = subprocess.Popen(["valgrind", "--tool=callgrind"] + command)
+    time.sleep(1.0)
+  else:
+    pid = subprocess.Popen(command)
+    time.sleep(0.1)
+  return plasma_store_name, pid
+
+def start_plasma_manager(store_name, redis_address, num_retries=20, use_valgrind=False, run_profiler=False):
   """Start a plasma manager and return the ports it listens on.
 
   Args:
     store_name (str): The name of the plasma store socket.
-    manager_name (str): The name of the plasma manager socket.
     redis_address (str): The address of the Redis server.
     use_valgrind (bool): True if the Plasma manager should be started inside of
       valgrind and False otherwise.
 
   Returns:
-    The process ID of the Plasma manager and the port that the manager is
-      listening on.
+    A tuple of the Plasma manager socket name, the process ID of the Plasma
+      manager process, and the port that the manager is listening on.
 
   Raises:
     Exception: An exception is raised if the manager could not be started.
   """
   plasma_manager_executable = os.path.join(os.path.abspath(os.path.dirname(__file__)), "../../build/plasma_manager")
+  plasma_manager_name = "/tmp/scheduler{}".format(random_name())
   port = None
   process = None
   counter = 0
@@ -245,7 +279,7 @@ def start_plasma_manager(store_name, manager_name, redis_address, num_retries=20
     port = random.randint(10000, 65535)
     command = [plasma_manager_executable,
                "-s", store_name,
-               "-m", manager_name,
+               "-m", plasma_manager_name,
                "-h", "127.0.0.1",
                "-p", str(port),
                "-r", redis_address]
@@ -260,6 +294,6 @@ def start_plasma_manager(store_name, manager_name, redis_address, num_retries=20
     time.sleep(0.1)
     # See if the process has terminated
     if process.poll() == None:
-      return process, port
+      return plasma_manager_name, process, port
     counter += 1
   raise Exception("Couldn't start plasma manager.")
diff --git a/src/plasma/plasma_manager.c b/src/plasma/plasma_manager.c
index 62ceee7be..2c82b8dc9 100644
--- a/src/plasma/plasma_manager.c
+++ b/src/plasma/plasma_manager.c
@@ -28,6 +28,7 @@
 #include "utstring.h"
 #include "common.h"
 #include "io.h"
+#include "net.h"
 #include "event_loop.h"
 #include "plasma.h"
 #include "plasma_client.h"
@@ -231,17 +232,6 @@ void remove_object_connection(client_connection *client_conn,
   free_client_object_connection(object_conn);
 }
 
-/* Helper function to parse a string of the form <IP address>:<port> into the
- * given ip_addr and port pointers. The ip_addr buffer must already be
- * allocated. */
-/* TODO(swang): Move this function to Ray common. */
-void parse_ip_addr_port(const char *ip_addr_port, char *ip_addr, int *port) {
-  char port_str[6];
-  int parsed = sscanf(ip_addr_port, "%15[0-9.]:%5[0-9]", ip_addr, port_str);
-  CHECK(parsed == 2);
-  *port = atoi(port_str);
-}
-
 plasma_manager_state *init_plasma_manager_state(const char *store_socket_name,
                                                 const char *manager_addr,
                                                 int manager_port,
@@ -257,8 +247,6 @@ plasma_manager_state *init_plasma_manager_state(const char *store_socket_name,
     state->db = db_connect(db_addr, db_port, "plasma_manager", manager_addr,
                            manager_port);
     db_attach(state->db, state->loop);
-    LOG_DEBUG("Connected to db at %s:%d, assigned client ID %d", db_addr,
-              db_port, get_client_id(state->db));
   } else {
     state->db = NULL;
     LOG_DEBUG("No db connection specified");
@@ -362,8 +350,6 @@ void send_queued_request(event_loop *loop,
   plasma_request manager_req = make_plasma_request(buf->object_id);
   switch (buf->type) {
   case PLASMA_TRANSFER:
-    LOG_DEBUG("Requesting transfer on DB client %d",
-              get_client_id(conn->manager_state->db));
     memcpy(manager_req.addr, conn->manager_state->addr,
            sizeof(manager_req.addr));
     manager_req.port = conn->manager_state->port;
@@ -460,8 +446,6 @@ client_connection *get_manager_connection(plasma_manager_state *state,
   client_connection *manager_conn;
   HASH_FIND(manager_hh, state->manager_connections, utstring_body(ip_addr_port),
             utstring_len(ip_addr_port), manager_conn);
-  LOG_DEBUG("Getting manager connection to %s on DB client %d",
-            utstring_body(ip_addr_port), get_client_id(state->db));
   if (!manager_conn) {
     /* If we don't already have a connection to this manager, start one. */
     int fd = plasma_manager_connect(ip_addr, port);
@@ -839,9 +823,7 @@ void process_message(event_loop *loop,
     process_wait_request(conn, req->num_object_ids, req->object_ids,
                          req->timeout, req->num_returns);
     break;
-  case PLASMA_SEAL:
-    LOG_DEBUG("Publishing to object table from DB client %d.",
-              get_client_id(conn->manager_state->db));
+  case PLASMA_SEAL: {
     /* TODO(swang): Log the error if we fail to add the object, and possibly
      * retry later? */
     retry_info retry = {
@@ -851,7 +833,7 @@ void process_message(event_loop *loop,
     };
     object_table_add(conn->manager_state->db, req->object_ids[0], &retry, NULL,
                      NULL);
-    break;
+  } break;
   case DISCONNECT_CLIENT: {
     LOG_INFO("Disconnecting client on fd %d", client_sock);
     /* TODO(swang): Check if this connection was to a plasma manager. If so,
diff --git a/src/plasma/test/test.py b/src/plasma/test/test.py
index c721ef143..719474d06 100644
--- a/src/plasma/test/test.py
+++ b/src/plasma/test/test.py
@@ -1,17 +1,17 @@
 from __future__ import print_function
 
+import numpy as np
 import os
+import random
 import signal
 import socket
 import struct
 import subprocess
 import sys
-import unittest
-import random
-import time
 import tempfile
 import threading
-import numpy as np
+import time
+import unittest
 
 import plasma
 
@@ -63,19 +63,12 @@ if not os.path.exists(redis_path):
 class TestPlasmaClient(unittest.TestCase):
 
   def setUp(self):
-    # Start Plasma.
-    plasma_store_executable = os.path.join(os.path.abspath(os.path.dirname(__file__)), "../build/plasma_store")
-    store_name = "/tmp/store{}".format(random.randint(0, 10000))
-    command = [plasma_store_executable, "-s", store_name, "-m", str(PLASMA_STORE_MEMORY)]
-    if USE_VALGRIND:
-      self.p = subprocess.Popen(["valgrind", "--track-origins=yes", "--leak-check=full", "--show-leak-kinds=all", "--error-exitcode=1"] + command)
-      time.sleep(2.0)
-    else:
-      self.p = subprocess.Popen(command)
+    # Start Plasma store.
+    plasma_store_name, self.p = plasma.start_plasma_store(use_valgrind=USE_VALGRIND)
     # Connect to Plasma.
-    self.plasma_client = plasma.PlasmaClient(store_name, None, 64)
+    self.plasma_client = plasma.PlasmaClient(plasma_store_name, None, 64)
     # For the eviction test
-    self.plasma_client2 = plasma.PlasmaClient(store_name, None, 0)
+    self.plasma_client2 = plasma.PlasmaClient(plasma_store_name, None, 0)
 
   def tearDown(self):
     # Kill the plasma store process.
@@ -258,21 +251,8 @@ class TestPlasmaManager(unittest.TestCase):
 
   def setUp(self):
     # Start two PlasmaStores.
-    plasma_store_executable = os.path.join(os.path.abspath(os.path.dirname(__file__)), "../build/plasma_store")
-    store_name1 = "/tmp/store{}".format(random.randint(0, 10000))
-    store_name2 = "/tmp/store{}".format(random.randint(0, 10000))
-    manager_name1 = "/tmp/manager{}".format(random.randint(0, 10000))
-    manager_name2 = "/tmp/manager{}".format(random.randint(0, 10000))
-    plasma_store_command1 = [plasma_store_executable, "-s", store_name1, "-m", str(PLASMA_STORE_MEMORY)]
-    plasma_store_command2 = [plasma_store_executable, "-s", store_name2, "-m", str(PLASMA_STORE_MEMORY)]
-
-    if USE_VALGRIND:
-      self.p2 = subprocess.Popen(["valgrind", "--track-origins=yes", "--leak-check=full", "--show-leak-kinds=all", "--error-exitcode=1"] + plasma_store_command1)
-      self.p3 = subprocess.Popen(["valgrind", "--track-origins=yes", "--leak-check=full", "--show-leak-kinds=all", "--error-exitcode=1"] + plasma_store_command2)
-    else:
-      self.p2 = subprocess.Popen(plasma_store_command1)
-      self.p3 = subprocess.Popen(plasma_store_command2)
-
+    store_name1, self.p2 = plasma.start_plasma_store(use_valgrind=USE_VALGRIND)
+    store_name2, self.p3 = plasma.start_plasma_store(use_valgrind=USE_VALGRIND)
     # Start a Redis server.
     redis_path = os.path.join(os.path.abspath(os.path.dirname(__file__)), "../../common/thirdparty/redis-3.2.3/src/redis-server")
     redis_port = 6379
@@ -281,12 +261,10 @@ class TestPlasmaManager(unittest.TestCase):
                                              "--port", str(redis_port)],
                                              stdout=FNULL)
     time.sleep(0.1)
-
     # Start two PlasmaManagers.
     redis_address = "{}:{}".format("127.0.0.1", redis_port)
-    self.p4, self.port1 = plasma.start_plasma_manager(store_name1, manager_name1, redis_address, use_valgrind=USE_VALGRIND)
-    self.p5, self.port2 = plasma.start_plasma_manager(store_name2, manager_name2, redis_address, use_valgrind=USE_VALGRIND)
-
+    manager_name1, self.p4, self.port1 = plasma.start_plasma_manager(store_name1, redis_address, use_valgrind=USE_VALGRIND)
+    manager_name2, self.p5, self.port2 = plasma.start_plasma_manager(store_name2, redis_address, use_valgrind=USE_VALGRIND)
     # Connect two PlasmaClients.
     self.client1 = plasma.PlasmaClient(store_name1, manager_name1)
     self.client2 = plasma.PlasmaClient(store_name2, manager_name2)
diff --git a/test/runtest.py b/test/runtest.py
index e632a325c..be97449c1 100644
--- a/test/runtest.py
+++ b/test/runtest.py
@@ -268,7 +268,7 @@ class APITest(unittest.TestCase):
       return x + 10
     while True:
       val = ray.get(f.remote(0))
-      self.assertTrue((val == 10) or (val == 1))
+      self.assertTrue(val in [1, 10])
       if val == 10:
         break
       else: