Legacy HDS Forums

TrueCopy initial sync traffic can cause performance issues

Discussion created by Legacy HDS Forums on Jan 10, 2008
Latest reply on Jan 16, 2008 by Legacy HDS Forums

Originally posted by: snowyx5

Here is the scenario:

We set up some new LUNS and put them in a TrueCopy relationship.
The LUNS are Serial ATA(groan) in an AMS500 virtualised through USPs (presented as external storage via the USP)
Configuration is identical between sites (yes, really, 100%)
The TrueCopy links between our USPs are pretty wide in terms of bandwidth.
We do a paircreate with the copy pace to 15 (the highest).
We do one lun at a time.

Here is what happens:

Write pending % on the SVOL side USP ramps up immediately.
It gets to about 70% and stabilises and remains that high until TC sync completed.
Performance on the PVOL side takes a serious dive for any hosts with other TrueCopy luns presented.
It seems that the SVOL side USP cannot destage rapidly enough and there is a knock effect to the PVOL side.

So, now we only do one lun at a time and set copy pace to 1 (the lowest).
Now the cache WP climbs to 60% and remains stable while the sync occurs.
Performance remains acceptable on the PVOL USP.

Now the query - prick the brains of the Forum ... :-)
Is there any way of controlling [i]initial sync [/i] TrueCopy traffic on the arrays ?
I'd like to deprioritise it even more than the copy pace allows.
I'd like to stop it from taking over the array when I'm careless.
Am I asking for too much, should we accept these are SATA drives and just manage them accordingly ?