BCV Full Establish affect on DMX3 WP limit

Question

We are experiencing performance issues on a DMX3. It has maxed out cache and disk. While I am aware that timefinder operations actually emulate symclone, hosts are using the symmir commands with BCV devices to sync and split. When incremental establishes are initiated, WP looks OK, but if a full est is kicked off, preformance on several business critical applications suffers. We are in the process of trying to schedule certian BCV ops to after business hours, but it has been tough. I have two related questions:

1 - I seem to remember in a timefinder course long ago learning how timefinder ops affected cache and WP limits...but could be wrong, of course. Is it true that only a portion of cache can be used for timefinder, and when a full establish is kicked off the majority of power goes toward completing the full, while any current syncs will see WP increase?

2 - How does symclone compare to timefinder as far as performance goes? We have at least a dozen different servers that require daily incr establish and split, and many others that will complete a full establish every few days.

Thanks for any insight and help you can provide!!

Quincy561 · Answer

There is one big difference between true BCVs and Clones. BCV devices become mirrors of the source device, occupying a mirror position. Because of this fact, you cannot fence the target devices off into a cache partition like you can with clones.

Both types of local copies will however generate WP tracks while copying. This can impact performance, which it seem it is in your case.

Have you tried applying a copy QoS value on the source devices? This could slow the copy down, therefore reducing the impact on your workloads.

ksmith123 · Answer

We are thinking about QoS, but unfortunately, even more critical apps/DBs depend on the timeliness of the BCV sync. We haven't quite figured out how to apply it at this time.

My main concerns are either validating or disputing my memory of how BCV affects the frame, and whether clone is actually a less intensive process.

Example - we did a full establish on 16 new 45 GB hypers during off hours. WP for those specific hypers wasn't all that bad, but the affect on the 40 top WP producing BCVs was 50x to 100x during the full establish. This is typical activity, even if performance issues aren't noticed?

Does the equivalant of a full establish on a clone result in the same performance hit as a full on BCV?

Quincy561 · Answer

Overall the impact can be similar. However if your BCVs are unprotected, and the clone targets are RAID5, the impact from clones for the same number of devices being established can be less for the unprotected BCV (less work to destage)

ksmith123 · Answer

Can you point me to a good doc on symclone - one that includes performance information?And can you confirm or deny my theory on how symmir full est affects the fram (a couple of comments above)?

Quincy561 · Answer

If you switch to clone emulation, you can split before the process says 'synchronized', if the time for the copy to be ready is critical.

Symmetrix

BCV Full Establish affect on DMX3 WP limit

Was this post helpful?