Best options for offiste backups

[MGT1981]

Hi everyone,

I have a relatively small client (40 or so users) that has a fairly large (~50 or so TB) backup set that has constantly changing data. They are currently running Veeam backups to an on site Exagrid and it is working flawlessly. We however need to get the backups sent off site somehow. The way I see it, we have 3 options.

1) Copy the backups from the exagrid to tape, the client uses an LTO-5 autoloader, it would obviously take several tapes and quite a while but it could be done. There is one problem with this however, with the Exagrid we are using the dedicated data mover share for Veeam, because of this I don't believe you can copy the data from the Exagrid to tape so I would need to hit the source twice. I would like to avoid this if possible.

2) Copy the data to a veeam cloud connect provider. The client has a very large internet pipe so I am not so much concerned about that, however backing up that much data to the cloud is going to be a huge expense.

3) purchase another Exagrid and stick it in a colo and replicate it. This option would work however the scalability gets very expensive very quickly.

I was just curious to get other people's thoughts here.

[Gostev]

Hi, what off-site data retention requirements are we talking about here? Months, years?

[MGT1981]

It varies by dataset but there is a good portion that will likely require a 24 month retention

[Gostev]

I personally would go with option (1) then. You can certainly do ExaGrid to tape with Veeam tape jobs, no issues there. It's slightly more expensive option in terms of on-going costs (due to having to physically deal with tapes), but scales pretty much infinitely at little extra costs. And super reliable (assuming proper tape handling).

May be (2) as well, if you can cut a good deal with some Service Provider. While your data foot print is huge, smart provider can achieve very low overall costs by not keeping GFS points on disk, and instead dumping them to tape. I am sure your customers don't really need all 2 years worth of data available online at any time. Perhaps only last 4 weeks, which typically covers 95% of restore scenarios.

Definitely not (3), as with storage based replication your backups will remain in the same fault domain (you want your second copy to be independent).

[bernardw]

Are you looking for the off site data storage as an archive point or a disaster recovery point? (Basically are you keeping your 2 years worth of data on the current Exagrid and just looking for a location for disaster recovery)?

Another question would be - how much data changes from day to day?

Personally, since I am trying to figure out how to get my data off site (mostly for disaster recovery purposes), I have been using LTO 3 tapes for 4 TB of data and am working on trying to move to rotated hard drives. I am really leaning towards the best option being having a Veeam repository at a remote location and using a backup copy job to move data over to there. I think this does not fit in Gostev's issue with your third solution but I am no where near as well versed on the subject as he is.

[Gostev]

I think this does not fit in Gostev's issue with your third solution but I am no where near as well versed on the subject as he is.

Backup Copy job decouples source and target storage devices from each other. First, it transfer the entire restore points (not just changed block on storage LUN level), and validates the VM data in this process. Second, it provides periodic health-check of the latest VM state to catch silent storage corruptions on target (and automatically fixes any corruption found). And finally, it stores archived full backups (aka GFS restore points) as an independent fulls which are completely not dependent on each other (so losing certain common data block only corrupts a single archived full, instead of all of them - as is the case with forever-incremental chains).

I agree what you are thinking is the best solution for your specific scenario, considering your environment size is fairly small. Tape really shines in large environments with high churn rate and long offsite retention requirements. In every other case, Backup Copy to a remote repository will be a better solution (as it provides for faster restores from off-site backup, and lower TCO).

[MGT1981]

Thanks for the response guys

I do have a question however, regarding backup copy jobs. A year or so ago I called support because of an issue with a backup copy job and the person I spoke to suggested that they not be used due to some problems Veeam was having with them. He instead suggested that we create 2 different jobs, on site using a weekly full and a nightly reverse incremental and a separate offsite job with a weekly full and a nightly forward incremental. I am wondering has this recommendation changed?

[foggy]

Most likely this was a recommendation in your particular situation as there was never a general recommendation for not using backup copy jobs at all (several issues could prevent or limit some specific scenarios, though). Hard to speculate without knowing the case details, however such issues are typically addressed by the very next patch/update, so it stands to try with the latest version.

[veremin]

The recommendations given sound strange to me. The root cause was not found, and just a workaround was proposed.

Also, since that time a lot of improvements and enhancements have been introduced to backup copy jobs. So, it still is and has always been a recommended way to create secondary copy of your backup data.

Thanks.

[TerryA]

I'm also trying to get my backups offsite as well and running into issues.

I have two Exagrid appliances, one at my main site and one at our DR location offsite. We do Veeam backup jobs Monday - Friday evening which utilizes the Exagrid Veeam Accelerated Data Mover. For the other Exagrid server I use a CIFS share and send Backup Copy jobs there. We have a 30Mb pipe between sites. I've been told to keep the last months worth of backups offsite. I keep the last 25 restore points (gets me a little over a month) at our DR site but I'm unable to get all of the backup copy jobs to replicate in the 24 hour period. I don't have a lot of change rate for about 50-60 servers (around 13TB total). I've tried using a WAN accelerator (although the WAN volume isn't on SSD) and without and not getting the results I want. Right now I'm just doing backup copy jobs for some of our most critical servers that I can fit into that window.

Any tips would be appreciated.

[veremin]

What is identified as a major bottleneck? And what transfer rates you get? If a backup copy job can't create a restore point within the specified backup window, you should either increase the backup window or improve job performance (working with bottlenecks) to make it process data faster. Thanks.