Continuous Access Performance Questions

[Paul Lessor]

We are attempting to move the entire contents of one EVA 8000 to another using Continuous Access (problems with first EVA).

We have the two EVA's plugged into two fabrics with even ports in A fabric and odd ports in B fabric.

We have a number of large Virtual Disks ranging in size from 500 GB to 1000 GB.

CA is VERY slow in creating replicas. It seems that each controller will only open up one connection to the other controller on each fabric. Additionally we are seeing throughputs of only 20 to 40 MB/s. This is horrendous. Our switches are Cisco MDS 9216i and MDS 9124 on fabrics A and B.

We see that the initial replication is slow when the destination array is allocating the space, but it seems to never speed up. We have tried to nudge it out of the self-imposed throttling mechanism by suspending and resuming replication and this seems to work, but I wish someone somewhere could better explain how best to go about moving so much data between arrays. HP support has not been much help. We've had to instruct them on a few issues.

Any help is appreciated.

Thank you,

[Ivan Ferreira]

>>> We've had to instruct them on a few issues.

Yes. It happens.

I suppose that the replication process is designed to no impact the performance of the normal data processing, and you don't have a tunable option to increase the speed (like rebuild_priority in some arrays for data reconstruction).