Add Arch docs

docs/_config.yml

@@ -90,6 +90,8 @@ sidebar:
           url: /installation-from-docker/
         - title: Installation from Source
           url: /installation-from-source/
+        - title: Planning your Talk Architecture
+          url: /planning-architecture/
     - title: Configuration
       children:
         - title: Required Configuration

docs/source/01-03-installation-from-source.md

@@ -74,4 +74,4 @@ yarn watch:server
 ```
 
 At this stage, you should refer to the [configuration](./configuration/) for
-configuration variables that are specific to your installation.
+configuration variables that are specific to your installation.

docs/source/01-04-planning-architecture.md

@@ -0,0 +1,21 @@
+---
+title: Planning your Talk Architecture
+permalink: /planning-architecture/
+---
+
+Talk is architected to be able to run on as little as 500MB of RAM. To do this however you will need to use the pre-compiled Docker container, as compiling the code and dependencies will cause a memory spike.
+
+For the average small blog or newsroom, these are our recommended machines:
+
+*Google Cloud*: g1 small, ~$14/month
+*AWS*: t2small, ~$16/month
+*Digital Ocean*: ~$10/month for their 1GB droplet
+
+From there, you’re free to separate app servers and DB servers, and scale up as much as you need.
+
+One larger newsroom’s setup, as an example of Talk performing at scale, is:
+
+Application servers: c4.xlarge (16 VM nginx + Talk VM machine pairs)
+Mongo nodes: 3x c3.medium (large db cluster, 1 master, 2 read replicas)
+
+If you need help with Talk performance or want custom scaling help or recommendations, let us know by logging a ticket and one of our engineers will get in touch with you: support.coralproject.net