Replication and scalability
Mon, 03/21/2011 - 11:30
We are currently evaluating couchdb for use in a distributed application. It involves a three tier architecture, with clients running an instance of couchdb, each of which replicates to a tenant couchdb server, each of which then replicates to the corporate couchdb cluster.
At peak adoption, on average 10 clients would sync to each tenant server, and 100,000 tenant servers would sync to the corporate cluster. The corporate cluster needs to be able to handle about 2000 writes per second and about 4000 reads per second.
Can couchdb handle this kind of volume?
Michael
Michael,
It's an unusual scenario, and at the moment we don't have performance numbers for the combined Couchbase cluster product which is in the works. Based on current membase cluster performance, that volume is within the ballpark. 6K gets+sets/second is feasible with predictable low latency, as long as you have enough RAM in the cluster to hold your working set of data.
I will alert our product management team to your needs, and perhaps we can arrange a phone conversation to better understand the parameters of your project and how we might be able to serve you best. I'd appreciate it if you would contact me at tim@couchbase.com with contact info to make that happen.
Regards,
Tim