disaster strategy/restore performance

[theviking84]

I believe with v10a veeam further enhanced the object storage interaction so that restore speeds improved more.

I read this article by Anthony Spiteri. https://anthonyspiteri.net/recovery-spe ... t-storage/ One of the key points though, is that he was restoring from an Amazon S3 bucket into a vmware cloud (on aws) infrastructure. I do have that, and can restore into that and get very good speeds as he did.

But, some of our infrastructure is on premise and in various locations. I've already made it clear to our management that we need replication for any HUGE disaster so that the vms are all spun up somewhere and we are NOT relying on restores themselves.

But, I still will have to wait until that is decided on.

So my questions here are,

1. If someone has 1gb internet, and needs to restore entire vms, dozens of TB worth, from an Amazon S3 bucket back to on prem vmware infrastructure, is this something that can be expected to be half decent on speeds?

2. What is veeam's stance on this kind of thing, because it seems to me like you have pushed more to get people using object storage and sobr/s3 and away from tape, as this way the space is way less, way easier to do it all, which I love, but then in the end tapes would seem to restore faster when needing to restore a lot of vms at once. I just want to understand if Veeam envisions customers relying on large multi vm restores from object storage and that this won't be a big problem for RTO. Or if they just figure you will be doing replication also and not trying to do that in a DR scenario.

Another thought, is I understand v10 has improved instant recovery engine and you can do multiple vm instant recovery from cloud amazon s3 bucket. Maybe this would be the choice, just run off the vms immediately while they are all copying their storage back to on prem over time. I wonder about performance there though and how feasible that really is.

Just trying to plan for the worst cases and this hit me today as a thing to think about.

[dalbertson]

1. Yes, you should see decent speeds across this pipe, depending upon how many you do at once
2. Veeams stance on this is that use what makes sense in your environment and RPO/RTO needs. Object is going to be better, Tape will continue to be around. Find the best use case for your situation.

Its possible to muliple IVMR, but remember they will be streaming from S3 and across the web. So while it is possible it will incur API / egress / bandwidth charges during the process. So unless its a tier1 app it may make more sense to restore instead.

One recommendation is that you can try to restore across the pipe, try IVMR, and also try download to performance tier - then restore to see which method is best for you.

[theviking84]

Thank you for that quick answer. I appreciate it.

I think overall I am going to try to still push that we replicate vms so that the object storage buckets will only ever handle like "restore files/folders, and mostly old ones" and everything else is performance tier, and then in true disaster where only bucket is there, we just use the replicas instead of even restoring.

I didn't think of moving from capacity to performance first and then restore. That's something to test. Thanks!

[dalbertson]

Sure thing. If anything you could tier out your instances....Tier 1 apps you would replicate and tier 2-3 apps you would be find performing a restore of.

[theviking84]

That's an even better idea. If we can just get the most critical ones replicated at least, and then know that some less important and hopefully some of the larger ones are restores and wait. Thanks again!