Backup Copy Job

[nholman]

Hi guys

Need some advice really around backup copy jobs. I will outline the environment first then ask my question.

2 Well connected sites at 100Mmbps. Doing just normal file copies from site to site using windows we reach speeds of 9 - 11MB/s.
Source site is the primary site with all backups going to a HP dedupe.

Our new site has a Dell Dedupe Dr4100. This site will eventually become our primary site as we are moving to it.

I have setup some backup copy jobs to make use of the "secondary job" feature in v7. These backup copy jobs at the source site are pointing to the Dell DR4100 as its repository and using the proxy located at that site.
In addition i have selected to use the wan accelerator on each side with the default values.

Now when the backup copy job kicks in and its copying our traffic seems very "bursty" and never solid. Looking at our incoming traffic at the target site it seems to shoot up to the full 100mbps then all the way down to 0. Then up again. then down. it doesn't seem right and goes agsinst what we see in windows file copying.
My question is, is this by design? Is the wan accelerator making it do this or go slower? i would attach an image of our traffic but have no idea how!
any help is much appreciated!

[Gostev]

Yes, this is by design - source WAN accelerator shoots at target one with 1GB data packages, as soon as they are prepared (one by one).

Using SSD drive for WAN accelerator cache (at least on the target accelerator) will help our engine to prepare these packages up to 10 times faster comparing to hard disk based cache, and thus will reduce these gaps dramatically. Also, with the upcoming Patch #2, using multi-core configuration for the source WAN accelerator will help to further improve the data processing performance (from 30% on regular cache, to up to 3 times with SSD cache).

So basically, with SSD and multi-core CPU, depending on your workload, you should be able to get pretty close to saturating your 100Mbps pipe (with Patch #2 installed). But don't be too obsessed with the complete saturation, remember that even with 50Mbps saturated, we are talking about at least 1Gbps of raw data transfer speed (20x average data reduction).

[RMullis]

Gostev,

Any idea on release date for patch #2 yet?

Rick

[nholman]

Hi gostev

thanks for getting back to me

So without SSD drives for the wan accelerator cache folder i wont really see any improvement?

We have a dell hybrid equallogic array - what if i was to attach a guest iscsi volume to the target proxy and set the cache to this? would i see a good improvement?
What about increasing the accelrator from 5 to say 10 in the settings would that make much difference?

[Gostev]

@Rick this month.

@Nick, hybrid may work well if you have 10Gb Ethernet for iSCSI. One of our architects is actually using custom hybrid storage, and he had seen great performance improvements with the pre-release patch code... I am just not sure what connectivity he used (lab environment, could have been direct). But having SSD in the picture certainly makes all the difference. I do not really understand your 5/10 question though.

[nholman]

Hi gostev

The target environment is 10gb ethernet iscsi so this may be an option then?

Sorry i was asking whether increasing the streams in the wan accelerator would make any difference at all?

[nholman]

ideally of course i would take the target device locally to the source device however this is not geographically possible so doing our first initial seed is the only option hence trying to get as much speed out of it as possible

[Vitaliy S.]

Hi Nick,

1. Yes, 10 GB Ethernet iSCSI may work well, indeed.
2. Yes, increasing the number of upload streams should help you better saturate your 100Mbps WAN link.

Thanks!

[Gostev]

Normally, it only makes sense to increase the default value for upload streams if your WAN connection has a very high latency.

[nholman]

while i have you guys, would you know of any complications of copying the backups on a HP D2D Dedupe Device to a Dell D2D Dedupe device? Would it not "pick up" de-duplicated data and cause problems or would it work? just trying to think of quicker ways to transfer the backups so we can then seed them as a straight drag and drop is a lot quicker

[Vitaliy S.]

Do you mean how it would work if you decide to copy backup files from one dedupe device to another one? I believe data would be rehydrated during copy job and then deduped again by the Dell device. If you want make the process quicker, then straight drag & drop of the backup files and jobs seeding would definitely save you time.

[nholman]

Is seeding only supported using a replication job? Not really looking to do that as there is no vcentre at the DR site yet.
I created a standard backup job and pointing the target straight to the DR Dell Dedupe - its now sitting at 12MB/s network transfer constantly so its flowing quite nicely.
I assume this would get quicker on subsequent jobs once the initial backup files are there?

theres jus so many ways for Veeam to do this type of stuff its hard to select the option without trying everything.

[Gostev]

Seeding is support for backup, backup copy and replication jobs. There quite a few existing discussions about this on the forums, in case you need more info.

[nholman]

ive looked for seeding on a backup job but cant find anything - i must be looking for the wrong thing?
is seeding on a backup job just CBT?

[veremin]

i must be looking for the wrong thing?

Then, this topic is likely to help you.

is seeding on a backup job just CBT?

Could you elaborate on it, since I’m not sure whether I follow you that?

Thanks.

[foggy]

If you're looking for the regular backup job seeding, then this is the topic: Seeding a backup (not replication).