SOBR Design

[cjack03]

Hi All,

I have started to implement a SOBR and have a couple of questions for the wonderful people of this forum. We have two proxy servers each using direct SAN access and storage snapshots. Our proxies are also our repository servers. The storage for the backup repositories is FC SAN LUNs from an entry level SAN and a backup 3PAR. This is on a single site, although across two data centres.

If i create 3 extents, two for fulls and one for incrementals and present one of the full repos to each proxy and the incrmental to just one and then in my job settings allow the use of both proxies i will cause some of the backup traffic to go across the NICs of the proxies to the extent which is not local??

I guess what I'm asking is if I use Direct SAN access and want to use a SOBR all of the extents of the SOBR should be on one single proxy and that should be the proxy used in the job setting? If this is the case and my proxy lacks muscle how can I add a second proxy and give it access to the same SOBR?

Many thanks.
Chris

[dellock6]

hi Chris,
your assumption is correct, there will be traffic between the extents grouped in the same SOBR group. However, you don't really have to "present" the repos to both proxies, it's already done by SOBR.
If you have a good enough network between the two proxies, like a 10Gb network, I'd simply create one ore more repo on each proxy (depending essentially on how is the raid groups created on those repos), and group all of them into a single SOBR. Then, both proxies will be able to consume evenly the SOBR, and jobs would be run by both proxies with no difference, compared to today as probably you have jobs sticked to one of the two proxies.

If you add another proxy, it will consume the same SOBR via the common network they are all connected to.

Luca

[cjack03]

Thanks Luca!

Unfortunately today we dont have 10Gb between the proxies. Guess i should start looking for some $$$.

Chris

[dellock6]

Well, I've not said that 10Gb is mandatory. There are for sure small designs where 1Gb or multiple aggregated 1Gb links can fit, don't get me wrong.
But for sure, as in any distributed environment, on the long run the bottleneck could become the network, that was my concept.