Re: Compaction through put and compaction tasks

[kurt greaves]

Number of active tasks is controlled by concurrent_compactors yaml setting.
Recommendation is set to number of cpu cores you have.

Number of pending tasks is an estimate generated by Cassandra to achieve a
completely compacted state (i.e, there are no more possible compactions).
How this is estimated depends on the compaction strategies in use.

If you increase the compaction throughput it simply increases the rate at
which C* will compact, allocating more resources to reading/writing from
disk.

On 26 September 2017 at 08:20, Anshu Vajpayee 
wrote:

> Hello -
>
> Ihave very generic question regarding  compaction. How does cassandra
> internally generate the  number of comapction tasks?  How does it get
> affected with compaction throughput ?
>
> If we increase the  number of compaction  throughput, will the per second
> compaction  task increase for same kind of write load ?
>
> Finally whats your  recommendation for  compaction throughput for SSDs?
>
> --
> *​C*heers,*
> *Anshu *
>
>
>

Compaction through put and compaction tasks

[Anshu Vajpayee]

Hello -

Ihave very generic question regarding  compaction. How does cassandra
internally generate the  number of comapction tasks?  How does it get
affected with compaction throughput ?

If we increase the  number of compaction  throughput, will the per second
compaction  task increase for same kind of write load ?

Finally whats your  recommendation for  compaction throughput for SSDs?

-- 
*​C*heers,*
*Anshu *