Network World's New Data Center Strategies Newsletter, 07/18/06
Virtual machines are usually instantiated from an operating system and application “image” that is stored in a SAN or NAS. There are two ways to create these images: Either a physical server is converted to a virtual server, or the virtual server is built from scratch. A physical machine can be converted to a virtual machine, either manually or with a specialized conversion tool such as Leostream P>V Direct or Platespin PowerConvert.
Once an administrator creates and stores a virtual server image, it can be booted onto any physical server running virtualization software. The image is sometimes known as a “freeze-dried” server, whereas the process of booting it is “instantiating” or creating a “server instance.”
The recoverability advantage comes from this ability to instantiate a server by booting it directly off a SAN or NAS, which may be remote to the physical server where it is instantiated. This leads to two levels of recovery:
* Local recovery and maintenance - If the physical server becomes unavailable due to a hardware failure or has to be powered down or rebooted for maintenance, you can re-instantiate the virtual servers onto a different physical server. The second physical server may be a “standby” server or it may be actively hosting other virtual servers and have spare capacity for an additional virtual server. The two physical servers may be in the same blade-frame (if they are blade servers) or in the same rack or even be completely different brands or models of servers in different parts of the data center.
* Remote recovery and disaster recovery - The second server does not even have to be in the same data center. If two data centers share access to a SAN or NAS, then administrators can re-instantiate virtual machines in a different data center from where they were originally running.
The latter scenario creates some intriguing possibilities for disaster recovery. If a data center fails for whatever reason, as long as the SAN or NAS is still available in a second data center, IT engineers can recover all the virtual machines by re-instantiating them to the second data center.
In fact, the second data center does not even need to be as fully equipped as the first one; it could have fewer servers. If that is the case, you would re-instantiate the virtual machines at a higher ratio (density) of virtual-to-physical. Consequently, organizations that can tolerate a slight drop in performance can build a much cheaper secondary data center to handle temporary disruptions. Further, organizations can simplify the disaster-recovery process by concentrating on the availability of storage and not worrying as much about synchronization at the application layer.
Organizations may require additional tools to implement this, depending on how much interruption they can tolerate. For example, backup software or continuous data protection software creates up-to-date snapshots of the virtual machines, so when they are re-instantiated there is minimal loss of state. For real-time recovery of virtual machines, IT engineers can use live-migration tools such as VMware’s Vmotion.
Server virtualization is not just for consolidation. Other applications of VM technology, such as business continuity/disaster recovery described here, can be more risky and less mature to implement, but the benefits can also be substantial. Nemertes Research provides additional analysis of the risk and maturity of virtualization-based solutions as part of the Data Center benchmark report.
Questions or problems regarding this web site should be directed to firstname.lastname@example.org.
Copyright © 2008 Art Beckman. All rights reserved.
Last Modified: March 9, 2008