2

Do we have any in-memory kind of engine just like Couchbase provides an in-memory layer and all the data sits over in-memory and if the node reboots your data is gone. we don't want to use yedis for caching layer if it provides any.

sai
  • 165
  • 1
  • 10

1 Answers1

3

No, YugabyteDB doesn't have an in-memory only engine.

It is a distributed database with persistence/replication as some of its core-tenets. You can run it with replication factor of 1, but it is designed with persistence in mind. If a node restarts, the data will still be around, and the in-memory block cache will be warmed up on demand very quickly.

Having said that, YugabyteDB is designed for very good read performance (sub-millisec latencies) even when data set doesn't fit entirely in RAM.

References:

  • "Extending RocksDB for Speed & Scale" describes some of this work done in YugabyteDB's storage engine that's based on RocksDB.

  • This post describes performance measurements for a random read workload where the data set does not fit in RAM.