Add some design documentation for replication

docs/replication.rst

+Replication Architecture
+========================
+
+Motivation
+----------
+
+We'd like to be able to split some of the work that synapse does into multiple
+python processes. In theory multiple synapse processes could share a single
+postgresql database and we'd scale up by running more synapse processes.
+However much of synapse assumes that only one process is interacting with the
+database, both for assigning unique identifiers when inserting into tables,
+notifying components about new updates, and for invalidating its caches.
+
+So running multiple copies of the current code isn't an option. One way to
+run multiple processes would be to have a single writer process and multiple
+reader processes connected to the same database. In order to do this we'd need
+a way for the reader process to invalidate its in-memory caches when an update
+happens on the writer. One way to do this is for the writer to present an
+append-only log of updates which the readers can consume to invalidate their
+caches and to push updates to listening clients or pushers.
+
+Synapse already stores much of its data as an append-only log so that it can
+correctly respond to /sync requests so the amount of code changes needed to
+expose the append-only log to the readers should be fairly minimal.
+
+Architecture
+------------
+
+The Replication API
+~~~~~~~~~~~~~~~~~~~
+
+Synapse will optionally expose a long poll HTTP API for extracting updates. The
+API will have a similar shape to /sync in that clients provide tokens
+indicating where in the log they have reached and a timeout. The synapse server
+then either responds with updates immediately if it already has updates or it
+waits until the timeout for more updates. If the timeout expires and nothing
+happened then the server returns an empty response.
+
+However until the /sync API this replication API is returning synapse specific
+data rather than trying to implement a matrix specification. The replication
+results are returned as arrays of rows where the rows are mostly lifted
+directly from the database. This avoids unnecessary JSON parsing on the server
+and hopefully avoids an impedance mismatch between the data returned and the
+required updates to the datastore.
+
+This does not replicate all the database tables as many of the database tables
+are indexes that can be recovered from the contents of other tables.
+
+The format and parameters for the api are documented in
+``synapse/replication/resource.py``.
+
+
+The Slaved DataStore
+~~~~~~~~~~~~~~~~~~~~
+
+There are read-only version of the synapse storage layer in
+``synapse/replication/slave/storage`` that use the response of the replication
+API to invalidate their caches.