Puredisk compaction

[gc_bus]

We have two Puredisk SPA's each with seven content routers addressing a total of 138TB per SPA. We only use Puredisk for Netbackup PDDO backups.

Does anyone else find that when compaction runs on the content routers that backup throughput drops considerably and if so what have you done to avoid this.

We are running Puredisk 6.6.0.3 (plus EEB30) and when the CRQP jobs kick in twice a day, this stops the compaction whilst they run and performance vastly increases as a result.

Any comments?

Thanks.

GC

[teiva-boy]

This is an achilles heel to all dedupe appliances...  Compaction, garbage collection, data validation.  They all call it something else..  I don't recall in PD if you can change the schedule of the compaction and other maintenance tasks, but perhaps move them to a different time?

[gc_bus]

Thanks for your reply. It seems that when compaction was running, the time we saw poor throughput, it was releasing a large quantity of space back to /Storage on three of the seven content routers. Other than that it's all back to normal now :) 

[gc_bus]

Ok, further experiments later and I can see that if we manually stop compaction on all of our content routers in our primary Puredisk SPA, then our overnight PDDO backups run more than twice as fast. Going to experiment further as of course with space that needs compacting not being compacted this may build up to too high levels unless compaction can run during the day.

I'll probably end up cron'ing compacting to only run during the day when there are fewer backups running.