EMC announced VPLEX at EMC world however, it was hinted at during Getstalt IT’s Tech Field Day at least in its asynchronous mode. What does VPLEX do? What does it mean to the virtualization and cloud industries? These were the discussion within the Blogger’s lounge at EMC World.
What does it do?
VPLEX maintains cache concurrency between two geographically separated storage arrays up to 100kms apart. Over that distance and VPLEX changes to an asynchronous mode. This asynchronous mode is not yet available even so, VPLEX offers a degree of disaster recovery unheard of today for non-virtualized loads, but for virtualization loads, it provides improved business continuity.
What does this mean for the Virtualization Industry?
In effect, EMC has moved VMware VMotion and SVMotion functionality down into the hardware such that any hypervisor can make use of it. It does not need to be just VMware vSphere. Granted, with VMware vSphere you have quite a few new possibilities. Such as the ability to ‘teleport’ a datacenter from the path of an impending natural disaster such as a hurricane, etc. Granted this does take quite a bit of technology to make work.
The key issue is how to have the same IP address space between two sites. Cisco has this covered with their Overlay Transport Virtualization (OTV) which allows two Nexus 7K to stretch the Layer 2 network between them. OTV is a MAC routing scheme which encapsulated IP Packets within Ethernet frames. Since the IP Packets are encapsulated the switch on both sides needs to understand how to interpret the traffic, at the moment that means using two Nexus 7Ks as they are the only switches with OTV support.
VPLEX solves the issue of cache consistency between each site. In effect it is megacache that writes the blocks through to the storage device on each side maintaining synchronous storage arrays. In fact, VPLEX will work with any storage array and not just EMC’s arrays. Due to latency issues, the reads are cached locally but at the same time they check to ensure the blocks have not changed on the local array. This is very similar to how EMC FAST maintains the FAST Cache.
More to Do
There is more to do with VPLEX, specifically working to get asynchronous or on demand data written from one array to main arrays across the world. So instead of maintaining cache consistency within a pair of arrays, asynchronous VPLEX will allow you to quickly access data as needed from multiple remote sites. This could be a great use for remote office situations.
The VCE Coalition is working together to provide the necessary components to move virtualization forward. From Cisco is OTV, from EMC is VPLEX, and VMware provides the management. In essence, VPLEX is an enabling technology that will allow VCE to easier sell VBlocks. How? In effect, you could sell a VBlock with VPLEX as a drop in hot site with no fuss and unified management.
It would also be interesting to see VPLEX from the local site into cloud storage for on demand business continuity needs. Just in case there are those impending disasters.