WAN Accelerator

[chrisbarr35]

Hi

I have a few questions regarding enabling the WAN Accelerator for some existing backup copy jobs;

1. I enabled the WAN Accelerator for one of our existing copy jobs a few days ago, this copy job is still running after a few days (only at 31% so far - increments normally take a few hours to complete) and appears to be doing a full read of the VMs along with a task "Calculating disk x digests" for each VM. Is this normal? From reading the documentation the WAN Accelerator cache should be populated from the repository. How can the cache be "seeded" in the future to avoid having to do a full read of the VMs in each copy job when I enable the WAN accelerator?

2. I have purchased a couple of 200GB enterprise SSDs, what's the procedure for migrating the existing WAN Accelerator cache to the SSD?

3. It seems that a single SSD is the preferred storage for the cache rather than a RAID of SSDs, what happens if (or when!) the SSD fails? Can the cache just be copied from the opposite WAN Accelerator?

Thanks for your help!
Chris

[tsightler]

How fast is your link? If the increments were only taking a few hours are you sure it's worth using WAN accelerator? Have you installed the R2 patch? How big is the backup file you are attempting to copy?

If you're starting from a completely empty global cache on both sides, with seeded backup data, then yes, it requires calculating digests for the entire VM (this isn't actually transferred across the WAN, but still can take a long time). This isn't the same as the cache data. The source WAN accelerator contains digests data for each VM, the target WAN accelerator has a global cache.

1. I don't think there's anything technically wrong with using RAID with SSDs, I've only pointed out that using RAID0 is simply increasing the odds of a failure. If you're willing to run the in RAID1 then you'll get redundancy and faster read performance, but write I/O performance will be the same as a single drive at best. That's probably a reasonable trade if your concerned about redundancy.

2. That's also how you would move your cache to your SSDs, just stop the WAN Accelerator, copy the VeeamWAN folder to the new drive, remap the drive letters, start the WAN Accelerator service.

3. Since the source WAN accelerator contains the digests, and the target contains the global cache, you can't just copy from the "opposite" pair. Rebuilding a lost global cache on the target should only take a few extra hours compared to a normal run, however, losing the source cache, which contains the individual digests for each VM is actually much longer because Veeam has to rebuild the digests data which requires for the target WAN accelerator to read every block from the target repository for each VM. Veeam is able to calculate digests at ~120-150GB/hr on modest hardware but potentially faster as the hardware capabilities increase (mostly dependent on block retrieval from repository and CPU on target cache).

[chrisbarr35]

Thanks for the info, our link is 100Mbps between the DC and our DR site, we have several copy jobs that run simultaneously so the link gets saturated, my plan was to use the WAN Accelerator for some of the copy jobs to reduce the amount of data to be copied.

1. RAID 1 SSD looks like a better option then, I'll order two more!

2. The WAN Accelerator is currently running on servers with no spare HDD slots, I'm going to install a new server at both sites with the SSDs. How would I migrate the existing VeeamWAN folder to these new servers?

3. Considering how long calculating the digests for just one copy job is taking, I think it is very important to create a backup of the VeeamWAN folder if I can't just copy it from the other WAN Accelerator. Can I just; stop the service, then copy the folder and re-start the service when no copy jobs are in progress? How about restoring, would it matter if I had to restore a backup of VeeamWAN from a week ago?


Chris