HPE Storage Users Group

Hey guys - question for you all.

We're currently using Veeam to back up virtual machines to an HP D2D device in our VMWare environment. When Veeam v7 comes out, they will be releasing the capability to restore individual virtual machines from 3PAR snapshots.

In discussions here, some people have voiced the opinion that this should be our entire backup solution for virtual machines when it comes out. Essentially, we would use Veeam to fire off one snapshot of the datastore volumes each night, with a retention of 14 days. Then we would use 3PAR replication to replicate those volumes/snapshots offsite for disaster recovery purposes.

What are your guys' thoughts on this architecture? Is it ok to run off a hierarchy of snapshots like that indefinitely? I know 3PAR is good about it's snapshots, but I'm not sure how comfortable I am with that setup. I'm also not sure if that constitutes a solid backup strategy or not. Curious to hear what you guys think.

Steve

Is your concern is the performance impact that 14 snapshots, living off your production volume 24x7 will have?

If you know what your change rate, or write to read ratio, then we could get an idea of how concerned you should be. In a high change rate (lots of writes) workload, you would want to validate the setup to ensure the storage is sized to handle the overhead.

_________________
Richard Siemers
The views and opinions expressed are my own and do not necessarily reflect those of my employer.

Yeah, that is my concern exactly. I know the 3PAR system handles snapshots well, but I'm not sure how well. These are datastores for virtual machines running in VMWare so I'm guessing the change rate is fairly high.

My other concern is using snapshots as backups, but I suppose as long as the data is replicated off site it is essentially the same thing as a backup.

Hi.

For me - storage snapshots can not be the single and only backup.

A backup system should provide answers to many different scenarios, such as:

short and long term RPO - restore data from last hour, last day, but also for last year.

logical failure - what happens if an admin deletes a volume by mistake/purpose?

physical failure in disks, storage fabric, or whatever.

The backup cannot rely solely on the production storage itself - be it 3PAR with best configuration or other system. It can take advantage of snapshots, replication, raid and other features, but not as your single backup solution.

Yizhar

We are implementing a similar strategy. With 200+ TB you can not back it all up all the time. We will put backup agents on databases servers and file servers that have data that needs long term retention. We will take daily snapshots and retain them for 1-2 weeks. Your concern about living off snapshots is not really worded correctly. Your production will be running from the User CPG and the snapshots will be on the Copy CPG. the performance impact come as new data is written to the VM that is different than the data already there. The 3par uses Copy On Write, so as the inbound data comes in the 3par needs to copy the old data to the Copy CPG. 3par is very efficient in doing this so the hit should be very low. We have over 800 VMs and over 60 are SQL databases servers and all have snapshots. We have not had any performance impact. One thing I strongly recommend is to set up System Reporter, if you did not buy it with the array, buy it. It allows you to set up very detailed performance reports down to the volume level. Comes in handy when people complain that storage is slow and you can show them it is not.