Replace Storage Advice

[jlingg]

Hi,
I am currently backing up 9 ESX 4.1 hosts attached to two Equallogic PS 6000 SANS in on SAN group. Veeam is running on a Dell Poweredge 1950 with 3 Dell MD1000 direct attached storage arrays (Serial Attached SCSI). This is working fine except that the MD1000s are getting old and don't have a lot of storage (2TB, 2TB and 4TB). Would replacing these devices with some kind of NAS device result in a hit on performance? Is there any way to find out? Right now most of my jobs are showing Load stats like this (3/8/2012 4:29:33 PM :: Load: Source 40% > Proxy 55% > Network 89% > Target 97%)

Any ideas?

thanks
Jason

[Jfmoots]

The way your bottleneck stats are reading now, your slowest points during a job are your network and your target storage. If you replace the MD1000 with something "faster" and provide plenty of bandwidth between your proxy server(s) and the storage (repository server) you'll improve backup speeds.

I'm a fan of dedicated backup storage devices like Exagrid's and Data Domain's. In addition to being nicerealiable targets, they offer dedupe/compression and replication abilities. More drives typically results in higher speeds. For instance, if I need 2 TB of storage, I'd prefer 4 500 GB drives in place of 1 2TB drive. Knowing this, when I am sizing a target NAS box, I try to buy it with as many drives as I can. Smaller drives will cost less but you might actually wind up buying more than a single NAS box to meet your demand. Just food for thought...

[J1mbo]

Disk performance is key for synthentic transforms - a few SATA drives in RAID-5 will take WAY longer than (presumably) oodles of SAS drives in the MD's (personally though I do use a lot SATA RAID-10 which seems to work well). Also might be worth doing some monitoring on that old 1950 as the bottleneck may be it's CPUs as much as anything.

[Gostev]

The way your bottleneck stats are reading now, your slowest points during a job are your network and your target storage.

Actually the network is likely fine, but I suspect the product is missing the latest patch set

[jlingg]

You are correct - we are runing 6.0.0.158 - will download and apply patch now.

[Jfmoots]

You are correct - we are runing 6.0.0.158 - will download and apply patch now.

Now Gostev's head is going to grow even bigger

[jlingg]

Not sure I believe the difference in Stats after installing the update. Maybe some other factors involved?
3/9/2012 9:58:23 AM :: Load: Source 40% > Proxy 35% > Network 1% > Target 99%

[Gostev]

Stats are exactly the same as you had before, except the network load is now shown correctly. Previously, having the target overloaded affected network counter's value.

[dellock6]

So now, at least, you had a confirm that an upgrade of the backup storage will improve performances
I'm with James for dedup dedicated appliances (and a big fan of ExaGrid). If you count the deduplication ratio they offer, price per Gb at the end is lower than a NAS, and performances are usually great.