State backend considerations

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

State backend considerations

Nick Bendtner
Hi guys,
I have a few questions on state backends.
Is there a guideline on how big the state has to be where it makes sense to use RocksDB rather than FsStatebackend ? Is there an analysis on latency for a full checkpoint for FsSateBackend based on increase in state size ? 


Best,
Nick. 
Reply | Threaded
Open this post in threaded view
|

Re: State backend considerations

Arvid Heise-3
Hi Nick,

Both questions are hard to answer given that it depends on your hardware, access patterns (read/update), record size/structure, parallelism, and probably a ton of other parameters.

The usual approach is to simply evaluate it in your setting. Since it's a matter of configuration, you can do some A/B testing.

In general, you need RocksDB if you want to have incremental checkpoints, which is recommended if you have rather few updates and big state.

On Mon, Jun 22, 2020 at 2:13 AM Nick Bendtner <[hidden email]> wrote:
Hi guys,
I have a few questions on state backends.
Is there a guideline on how big the state has to be where it makes sense to use RocksDB rather than FsStatebackend ? Is there an analysis on latency for a full checkpoint for FsSateBackend based on increase in state size ? 


Best,
Nick. 


--

Arvid Heise | Senior Java Developer


Follow us @VervericaData

--

Join Flink Forward - The Apache Flink Conference

Stream Processing | Event Driven | Real Time

--

Ververica GmbH | Invalidenstrasse 115, 10115 Berlin, Germany

--

Ververica GmbH
Registered at Amtsgericht Charlottenburg: HRB 158244 B
Managing Directors: Timothy Alexander Steinert, Yip Park Tung Jason, Ji (Toni) Cheng