EMC is going to federate globally distributed storage arrays into a single system using VPLEX front-end virtualising boxes, and is presenting this as a big step towards private cloud computing.
Pat Gelsinger, EMC's chief operating officer and recent high-level and high-profile recruit from Intel, said: "VPLEX is ground-breaking technology [and] will enable follow the sun computing, the relocating of workloads to low-cost energy regions or moving them out of the way of approaching storms."
The concept of data at a distance is nothing if not ambitious. Take multiple geographically distant storage nodes and combine them into a logically single resource pool that provides fast access, secure access, and protects data against internal, silent corruption and disasters. No one else has done this, or even thinks it's practicable.
This won't all be achieved in one release - it's more of a journey than a finished product. Today EMC announced version one with VPLEX Local and VPLEX Metro products covering data centre and campus-wide distances. Future versions, VPLEX Geo and VPLEX Global for example, will add continental and global distance coverage enabling multiple data centres to be pooled into a single virtual resource, at least from the storage point of view.
EMC says that, ultimately, you will be able to move thousands of virtual machines and petabytes of information across thousands of miles and, say, move batch operations to locations with lower energy costs when practicable.
VPLEX combines scale-out clustering and caching with what it calls distributed cache coherence, which addresses latency, bandwidth and consistency issues, to keep widely separated arrays in a consistent state. EMC is trademarking this as AccessAnywhere and says it will eventually be added to more of its products.
The VPLEX hardware, sitting in-band between heterogeneous block (not file and not object) storage arrays and servers, consists of two VPLEX Directors in a high-availability setup, each of which has multi-core Xeon processors, 32GB of cache, and 8GBit/s Fibre Channel host and storage array connections.
VPLEX supports existing back-end EMC storage and third-party arrays, and has been qualified with, EMC says, "major hosts, clusters, operating systems, virtualisation platforms, [and] storage area network (SAN) infrastructure". An EMC staffer said: "VPLEX supports IBM's SAN Volume Controller (SVC) as a backend storage array, and all therefore all the storage SVC supports, today." Another person familiar with VPLEX said: "IBM will be thrilled. The heck they will."
A VPLEX Local cluster can have up to four VPLX engines, meaning eight directors, and supports up to 8,000 virtualised storage volumes, and provides "non-disruptive, transparent data mobility within, across and between MC and non-EMC storage within a single site/data centre".
VPLEX Metro can link two separate VPLEX clusters, with up to four engines each, within a data centre or up to 100km apart and federate some or all of the storage volumes across both clusters. There is a 5ms round-trip maximum response time limit. It supports up to 16,000 storage volumes and any volume can be configured for simultaneous access by applications in the two locations. The volume in one location is stretched as it were to cover the other.
VPLEX Geo, which will enable asynchronous federation across continental distances, and VPLEX Global, which will feature synchronous and asynchronous federation globally, should appear in 2011. VPLEX Local and Metro are available immediately. They can be bought with capacity-based licenses or as a private cloud-based "pay as you go" subscription license specific to VPLEX.
VPLEX Local starts from $77,000 while the subscription license starts from $26,000.
Here it is; EMC's big bet. HDS and NetApp have rushed out blogs saying we can federate already (HDS) or it's the wrong way to federate (NetApp). We will see how EMC's customers take to VPLEX and whether EMC and Gelsinger's audacious gamble pays off. ®