Cross region transactors and DynamoDB integration

[Vincent Tang]

I am doing research on using Datomic Transactor with my AWS DynamoDB. I
have DynamoDB tables and application instances in both AWS east and west
regions. My use case is, one UI application instance is doing the write
operations, it create/update/deletes data in DB. It only writes to DynamoDB
table in east region. Other service instances only read DB. I want to have
local cache in each of the service instances. I'd like the service
instances to get updates automatically whenever the UI application changes
the DB (I have another DynamoDB replication instance to replicate from east
to west). It seems Peer Library + Transactors is the best solution between
the applications and DynamoDB tables.



I'd like to have one Transactor in each of the regions to serve all service
instances in that region, for faster response. However I haven't learned
how the Transactor in west would know the DynamoDB changes if UI only
writes to the table in east. Will Transactor know if the DB changes are not
made through it? If yes, how does it do it?

[Linus Ericsson]

Have you tried to just connect your distant Peers with the transactor in
the west region (with all nescessary access to the backing storage)? As
long as the transactor is close to the backing storage everything should be
just fine for listening on changes that way. If you need higher read
performance in the distant application, could you just use local memcache
instances for that?

Having multiple transactors would not solve anything in this case,
especially not if the distant peers is just reading the database and
listening to transaction updates.

(I'm not sure on this, but I can't see how long distant Peers would slow
down the operation of the tighter cluster - the transactor notifies all
connected peers of successful transactions but down wait for connected
Peers to acknowledge or anything in any specific way).

/Linus