At EMCWorld today Pat Gelsinger had a pair of VPLEXes flanking him on stage and actively moving VMs from “Boston” to “Hopkinton” data centers. They showed a demo of moving a bunch of VMs from one to the other with all of them actively performing transaction processing. I have written about EMC’s vision in a prior blog post called Caching DaaD for Federated Data Centers.
I talked to an vSpecialist at the Blogging lounge afterwards and asked him where the data actually resided for the VMs that were moved. He said the data was synchronously replicated and actively being updated at both locations. They proceeded to long-distance teleport (Vmotion) 500 VMs from Boston to Hopkinton. After that completed, Chad Sakac powered down the ‘Boston’ VPLEX and everything in ‘Hopkinton’ continued to operate. All this was done on stage so Boston and Hopkinton data centers were possibly both located in the convention center but interesting nonetheless.
I asked the vSpecialist how they moved the IP address between the sites and he said they shared the same IP domain. I am no networking expert but I felt that moving the network addresses seemed the last problem to solve for long distance Vmotion. But, he said Cisco had solved this with their OTV (Open Transport Virtualization) for Nexus 7000 which could move IP addresses from one data center to another.
Later at the Expo, I talked with a Cisco rep who said they do this by encapsulating Layer 2 protocol messages into a Layer 3 packet. Once encapsulated it can be routed over anyone’s gear to the other site and as long as there was another Nexus 7K switch at the other site within the proper IP domain shared with the server targets for Vmotion then all works fine. Didn’t ask what happens if the primary Nexus 7K switch/site goes down but my guess is that the IP address movement would cease to work. But for active VM migration between two operational data centers it all seems to hang together. I asked Cisco if OTV was a formal standard TCP/IP protocol extension and he said he didn’t know. Which probably means that other switch vendors won’t support OTV.
There was a lot of other stuff at EMCWorld today and at the Expo.
- EMC’s Content Management & Archiving group was renamed Information Intelligence.
- EMC’s Backup Recovery Systems group was in force on the Expo floor with a big pavilion with Avamar, Networker and Data Domain present.
- EMC keynotes were mostly about the journey to the private cloud. VPLEX seemed to be crucial to this journey as EMC sees it.
- EMCWorld’s show floor was impressive. Lots of major partners were there RSA, VMware, IOmega, Atmos, VCE, Cisco, Microsoft, Brocade, Dell, CSC, STEC, Forsythe, Qlogic, Emulex and many others. Talked at length with Microsoft about SharePoint 2010. Still trying to figure that one out.
I would say the bloggers lounge was pretty busy for most of the day. Met a lot of bloggers there including StorageNerve (Devang Panchigar), BasRaaymon (Bas Raaymon), Kiwi_Si (Simon Seagrave), DeepStorage (Howard Marks), Wikibon (Dave Valente), and a whole bunch of others.
Well not sure what EMC has in store for day 2, but from my perspective it will be hard to beat day 1.
Full disclosure, I have written a white paper discussing VPLEX for EMC and work with EMC on a number of other projects as well.