Comments on: IBM DS4700 Model 70, VMware ESX 3.5 and Site Recovery Manager

By: Andre

Andre — Fri, 25 Sep 2009 00:46:29 +0000

G’day Dave,

I hope you and your other half are doing well.

This was definitely a very interesting project under some pretty tight timelines. Good on you for putting this out there. I only came across this site randomly looking for something else. It will definitely add value for anyone else looking to implement a similar solution as at the time saying doco was light is a bit of an understatement. Dave definitely knows his stuff guys.

Regards,

Andre

By: Dave V

Dave V — Tue, 21 Apr 2009 10:44:27 +0000

Sorry for the late responses on the above comments;

Erik, the customers requirements were to optimise the available space in the DR SAN for some other backup purposes. Agreed on your points, ideally, given the right environment, the best protection would be provided by using sync replication and matching RAID sets. The performance on the Prod side was of the most priority and therefore the decision was to use Async replication.

Imran, yes we had to break away from best practice and merge the fabrics. Unfortunate outcome, and would definately not recommend using a Model 70 to do SRM.

Thank you both for the comments!

By: Imran

Imran — Mon, 20 Apr 2009 13:52:28 +0000

Hi I should have read this before I got stuck with the same prblems. Can i ask what you did about problem number 1. I have a same situation where i have 1 port availble per controller connected to seperate switches. Do i have to merge the fabrics?

By: Erik Zandboer

Erik Zandboer — Fri, 20 Feb 2009 09:37:22 +0000

Hi,

You state that you have a mix of RAID10/RAID5 on the production site, and that you could consider to replicate to all RAID5 for storage optimization. But beware!!! If your replication is synchronous, you will loose performance (each write to RAID10 has to be committed to the RAID5 on the DR site as well before the write completes). Caching on the DR site will help some, but not in the end as cache fills up.

Asynchronous replication… I would stay away from it (I have experience with Sun x6140 which is basically the same unit I think), and asynchronous replication leaves you nowhere when the WAN “breaks” all of a sudden… (some blocks replicated, others not)… Not what you want when you implement for DR. Just my 2cts.