EVA Perf

[mark leibowitz]

We have an EVA8000 on our primary site and we are moving our EVA5000 to the DR site. We will be replicating between sites.Inorder to work out the bandwidth requirments between branches, we need to work out the amount of data changed at our primary site over 24hrs.How can you work this out. I have looked at evaperf but cannot find any suitable counters, they all relate to rate of data change, NOT the actualll amount that has changed, Any ideas?

[Ivan Ferreira]

Do you have this already replicating? You can use san switch commands like portperfshow to identify the ammount of data using the ISL port.

If you have business copy, you "may" create a demand allocated snapshot, and then, use the size of this snapshot to identify the ammount of changed data for a particular virtual disk.

[mark leibowitz]

The 8000 and 5000 are currently being replicated at the primary site. The 5000 will move to DR.
The problem i see with using snapshots is that you will only see the difference in data size between 2 points in time, however you will not pickup the data changes that happened inbetween.When u replicate between sites, it will replicate all changes including a "delete". I need the amount of data change and not the "growth"

I need to monitor continuously for 24 hrs, the amount and size of writes on each vdisk or disk group.

[Mark Poeschl_2]

I don't think getting a 24 hour total is what you'd want to size the 'pipe' between your sites. More than likely the volume of data changed varies quite a lot throughout the day. If this is so, you'll want to size your connection to handle the biggest volume of data changed over some much smaller interval. Set the interval according to how much risk you're willing to take of the DR site not being completely 'in sync.'

As for calculating the volume of change over any given interval see this quote from the EVAPerf Whitepaper:

"The counters are managed somewhat differently in HP StorageWorks Enterprise Virtual Array 3000 (EVA3000) and HP StorageWorks Enterprise Virtual Array 5000 (EVA5000) systems than in EVA4000, EVA6000, and EVA8000 systems. In EVA3000 and EVA5000 systems, most counters (Req/s, MB/s, and so on) are one-second averages per sample, regardless of the interval between samples. Therefore, each sample represents an instantaneous snapshot of the activity at that moment. In EVA4000, EVA6000, and EVA8000 systems, however, the counters are true averages over the sample interval, so highly variable data will have different characteristics for longer samples than for shorter."

So - once you select your interval - take the daily maxima of the sum of all Vdisk "Write MB/s" counters on your production EVA8000 as the volume of data transfer you'll need to support replication.

[mark leibowitz]

The best time to test this for max change would prob be month end.
I will test the evaperf with that counter and see what type of result i get

thanks

[Alzhy]

I can only think of probably a specialized software written to work with Continous Access/EVA for you to do this. If however you are using VxVM either VVR (Volume Replicator) or plane old simple FlashSnap Mirrors - it should be relatively easy...

Hey, EVA's now understand Active-Active. Why now switch to VxVM from LVM...?