HP StoreOnce Catalyst Cofc Slow

[slsc0728]

Hi ALL,
I have using HP StoreOnce 5100 Catalyst Cofc using 8gb x2 FC. and 8gb FC LTO7 Tape Library
But I found the Backup speed only around 100-150MB/s and
100MB/s for archive the backup to tape.

Anyone know how to increase the read/write speed of StoreOnce Catalyst?

Thanks!

[DGrinev]

Hi Dragon and welcome to the community!

Can you share the bottleneck stats, as it should help us to see the weakest link in the chain. Thanks!

[slsc0728]

Hi DGrinev,
I am using Veeam 9.5U3 and HPE StoreOnce is 3.16.3

Backup Job bottleneck stats
1/15/2018 10:02:19 PM :: Hard disk 2 (3.8 TB) 2.5 TB read at 162 MB/s [CBT]
1/16/2018 2:37:36 AM :: Busy: Source 38% > Proxy 53% > Network 11% > Target 99%

Backup to Tape Job bottleneck stats
2.5TB VBK file need 7 hours to finish.
The Process rate only 108MB/s
2/17/2018 6:04:17 PM :: Busy: Source 96% > Proxy 20% > Network 13% > Target 30%

Thanks

[DGrinev]

The bottleneck stats confirm the issue with the target storage speed, as 99% target means that the writer component spent most of the time performing I\O to backup files.
I'd recommend you to review the white paper "Veeam and Storeonce configuration guide", as it contains best practices. Thanks!

[slsc0728]

I have follow up this guide to setup storeonce .
But the speed is slow .

[DGrinev]

Hi,

You can contact the support team and let them take a closer look at the infrastructure, also they might initiate some tests against your environment and give helpful advises.
Please do not forget to put the case id in this thread, so we can track it internally. Thanks!

[foggy]

But the speed is slow .

Why do you consider the reported speed of 162 MB/s to be slow? It it the speed of the initial full backup to a dedupe appliance?

[slsc0728]

I am using 8gb FC connection, the MAX speed will be 800MB/s

But the Full Backup only got around 150MB/s , the incremental Backup only got around 80-200MB/s

It is Not a init Full, The StoreOnce have 2 Full Backup restore point.

[foggy]

You will not get close to what link allows with a single stream backup to a dedupe appliance. With several parallel jobs or per-VM chains enabled for repository (in case you're backing up multiple VMs), you could workaround the storage ingest rate limit of a single write stream and get better performance.

[splatt]

Hi I've been battling this for around 2 months with HPE and Veeam Tech support
One issue was that VM's with many hard disks get really slow e.g. a file server with 10 separate VMDK files can take hours, at 7MB/s per disk yet a VM with a single disk can transfer at >500MB/s over the same infrastructure - COFC 8GB FC

[jmely]

Is your Proxy and Catalyst gateway on the same server ? The Catalyst store in Low Bandwith ?

[davidburke84]

Are you running newer firmware on the SO? There have been some improvements in the newer version included number of open files IIRC. I'm running the the newest myself with no issues.

I would say your times are very similar to mine and I have a very similar setup as yours. You could try to change your jobs to land to the CIFS configuration on the SO, but you will not see the same dedup and you will not be able to do it over fiber. I tried to do this of 10GB network and was not pleased with the results.

There is a lot of extra processing that goes into the deduplicating the data. This is the trade off for backup performance. I'm currently in a bit of a redesign with my storage. I'm going to land my backups to local disk then run a copy job to our local SO. Are copying your data to a remote SO? I'm finding that doing a copy of the data from a local SO to remote SO through the Veeam software is getting worse compression. Veeam support recommended I change to land backups locally as well.

[davidburke84]

Hi I've been battling this for around 2 months with HPE and Veeam Tech support
One issue was that VM's with many hard disks get really slow e.g. a file server with 10 separate VMDK files can take hours, at 7MB/s per disk yet a VM with a single disk can transfer at >500MB/s over the same infrastructure - COFC 8GB FC

Never seen performance that bad. Do you have too many concurrent tasks enabled on your appliance?

[foggy]

One issue was that VM's with many hard disks get really slow e.g. a file server with 10 separate VMDK files can take hours, at 7MB/s per disk yet a VM with a single disk can transfer at >500MB/s over the same infrastructure - COFC 8GB FC

Basically, backup speed is expectedly higher if you backup multiple VMs, since each VM is processed within a separate write stream, while all disks of a single VM occupy a single stream, which is limited by ~150MB/s. That said, in your case the speed is indeed pretty slow (7MB/s).