by Hu Yoshida on Nov 29, 2007
I was recently on CIO Talk Radio with Sanjog Aul our host, Jeff Nick of EMC, John Webster of Illluminata, and Phil Edholm of Nortel Networks when the subject of Virtualization 2.0 came up. Virtualization 2.0 was introduced by IDC at their Virtualization Forum in February 2007. In October 2007 Gartner included Virtualization 2.0 in their top ten predictions for 2008.
So what is this all about? What differentiates Virtualization 1.0 from 2.0? Is it social networking like Web 2.0? In a way it is, if mobility is part of social networking.
First of all we need to understand that the virtualization that IDC and Gartner are talking about is Server Virtualization. Basically Server Virtualization 1.0 was all about server consolidation, the ability to stack applications along with their base operating systems (virtual machines) onto the same processor platform to increase utilization.. The difference in Server Virtualization 2.0 is the added dimension of virtual machine mobility and business continuity. Server Virtualization 2.0 would be represented by VMWare’s VMotion which enables virtual machines to move across physical processors, and VMWareHA (High Availability) that enables virtual machines to be restarted on another production server in the event of a server failure.
While virtualization numbering has not been applied to storage virtualization – yet, I believe there is a distinction to be made between the different levels of storage virtualization.
Storage Virtualization 1.0 would be the SAN based virtualization appliances that were introduced over 7 years ago to do volume virtualization. Their main value proposition was to provide consolidated volume management in the SAN which would eliminate the need for volume management licenses in each server. Storage virtualization 1.0 could create a volume that consisted of extents that spanned multiple storage arrays that were accessable on the SAN.
Similar to Server Virtualization 2.0, Storage Virtualization 2.0 added the dimensions of volume mobility and business continuance. This was introduced by the Hitachi Universal Storage Platform in Sept, 2007. The USP, with 128 internal processors, connected by a switch architecture to a large, dynamically configurable, global cache, had the horse power to do what a cluster of PC’s with separate 4GB caches could not. More importantly, as a control unit in the storage array, it had all the information about physical cache slots and track tables that were not available to an appliance sitting outside the storage arrays in a SAN. With this horse power and readily available information, the USP could move volumes to different tiers of storage, migrate volumes for technology refresh, and replicate volumes for business continuance, with out disruption to the application and without adding an additional burden to the SAN. Another characteristic of Storage Virtualization 2.0 is that it does not depend on the SAN. It can provide virtualization services to any server that connects through standard storage protocols, like ESCON, FICON, ISCSI, as well as FC. Through gateways it can also support various types of NAS, CAS, WEB, or email storage requirements.
So what is the next step for Storage Virtualization 3.0?
I believe the next step is capacity virtualization, or thin provisioning. This year Hitachi introduced the USP V which added Dynamic Provisioning on top of the capability to virtualize external storage. Dynamic Provisioning is the ability to provision physical storage when it is actually used, not when it is allocated. It eliminates the waste of allocated but unused capacity which some estimate to be 60% or more of the average storage allocation. It can also eliminate that waste from all the copies that are usually made for backup cycles, business continuance, data mining, development and test, and data distribution. This type of virtualization also enables a volume’s capacity to be automatically leveled across many spindles of RAID protected disks, which reduces management complexity and distributes the I/O load across many more physical disk arms in order to compensate for limitations of larger and slower disks.
If I were to number the levels of storage virtualization, it would go like this:
Storage Virtualization 1.0 SAN based appliance providing volume virtualization
Storage Virtualization 2.0 Control Unit based Storage Virtualization with volume mobility and business continuance
Storage Virtualization 3.0 Control Unit based storage virtualization with capacity virtualization or Dynamic Provisioning
If I were to edit Gartner’s prediction for the top 10 technologies for 2008, I would add Storage Virtualization 3.0. IBM and EMC have already stated that they will provide their versions of thin provisioning or Virtualization 3.0 in 2008.
Hu, following your 2.0 analogy with VMware, would it not be more appropriate to expect Storage Virtualisation 2.0 to provide the ability to move a logical LUN *between* storage arrays irrespective of the virtualisation component? For example, VMotion moves a virtual server between physical ESX servers; should Storage Virtualisation 2.0 not move a logical LUN between physical USPs?