Ephemeral bucket optimal/recommended count


I am using couchbase-server-community-5.1.1 server on centos6. I am having 10 ephemeral buckets, with 200MB bucket size for each. I have 3 node cluster with 1 replication factor. Cluster quota is 3000MB per node. I have one custom document type stored in all these buckets.
All these buckets are accessed by a group of services (SOA) with document access (key lookup).
I have around 15k-20k ops/sec hits on the cluster on peak load.

Looking to optimize the number of buckets as suggested in blogs

What would be the suggestion/recommendation for “count” of Ephemeral bucket with above specs. Should it be brought down to 5 or less? Since blog talk about Disk I/O which is not present in ephemeral buckets, should the count be brought down at all?

Any suggestions/recommendations will be appreciated.
Thanks in advance !!

Assuming you don’t have different replication factors, eviction policy, or require different permissions for each set of data, I’d probably just put all the data in a single bucket - and use key prefixing to distinguish between your different types / namespaces.

Note that while Ephemeral doesn’t have disks; there’s still management overhead (managing mutations, DCP replication etc) per bucket; so if you care about maximum efficiency the fewer buckets the better.

(As an aside, each bucket has a fixed memory overhead - I forget the exact numbers but you’ll see an empty ephemeral bucket still consumes ~ <10MB of memory. So if your bucket quota is only 200MB, you’re already wasting ~5% of your memory before you start!)

1 Like

Thank you :slight_smile: