Wondering how EMC’s storage federation device VPLEX might solve SharePoint DR concerns?
A few of the engineers teamed up to create a reference architecture that shows you how it’s done.
First, they listened to early customer feedback on what use cases might be most beneficial.
What came up again and again – SharePoint (and SAP and Oracle) DR is not that easy and virtualization of the servers helps a great deal, but the ability to perform VMotion across two sites without a major network upgrade is game-changing. If you weren’t aware of VMotion, it’s a feature that allows the migration of operational (live, running) guest virtual machines between similar but separate hardware hosts sharing the same storage. Each of these transitions is completely transparent to any users on the virtual machine at the time it is being migrated. Since VPLEX allows two sites to share the same storage, it doesn’t take a genius to realize that this enables VMotion across two sites.
Second, they built out the environment and tested a farm.
Specific to SharePoint, they configured the farm according to EMC and Microsoft best practices with about half a TB of total space. They used KnowledgeLake’s DocLoaderLite to populate SharePoint with random user data and then fired up a simulated load using Microsoft VSTS (Visual Studio). The Proven Solutions mantra held true – we wanted to get this as close to real world as possible and push the thresholds of performance and scalability at the same time. Our SharePoint guru and friend James Baldwin put a great amount of effort in putting this together amongst a team of several others (Don, Brahim, Haji, Patrick, Joe, Brian C, and many others).
Third, they provide the result and key findings.
I’d boil it down to this. VPLEX can safely be inserted in the data path between host and your existing storage array. You can stretch your clusters and enable a much better HA and DR strategy for your SharePoint farms. Failover across sites used to be one of the most challenging IT procedures out there… it’s why many companies don’t even have a DR plan. Now it can be as simple as a local failover, once it’s setup. And swapping out/upgrading storage becomes a seamless procedure compared to what might have happened yesterday. You can bounce your server back and forth without much of a blip.
OK, so where’s some detail on that “blip”?
Page 35 shows the VMotion durations, with and without latency added.
This section of the paper describes how the SharePoint farm response time will be affected before, after, and during the VMotion.
I think skeptics like Eddie would have liked to see data like this. Oh yeah, we did the same with non-SharePoint SQL databases (1 TB), Oracle E-Business Suite (with 11g), and SAP (ERP and BW). And it’s all in this little reference architecture here.