Datageek
Datageek

Reputation: 26679

Minimum Cassandra for a small dataset

I am trying to configure some really small Cassandra setup. The idea is that it should take as little resources as possible while performing as good as possible.

The current setup:

I know this setup is pretty terrible, however given the tiny datasets I am going to write here I was hoping for a decent performance.

The reading part is pretty good but writing is terribly slow. I am able to write only about 300 records per second (that is using the BEGIN/APPLY BATCH, which I would hope should be rather fast).

I am using a default Cassandra configuration. Is there any way to speed up the writing process? What would be the best way to reduce IO to speed things up?

Upvotes: 1

Views: 521

Answers (1)

oleksii
oleksii

Reputation: 35895

Is there any way to speed up the writing process?

With your set up, probably not. If the network file share is located on another PC with slow IOPS and low network speed then the write performance sounds reasonable (and actually good).

What would be the best way to reduce IO to speed things up?

You would want to increase IOPS, physically separate commit log from data folder and follow other recommendations. Get more memory, put drives on the local machine, add more machines to the cluster, but this may sound unrealistic with your constrains.

Maybe you can use something else with your hardware? In-memory database may suit you better, have a look at redis or memcached.

Upvotes: 1

Related Questions