When it comes to disaster recovery with virtualization, you get what you pay for. If your IT budget is too tight...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
to accommodate the most expensive VMware or Citrix packages, don't expect continuous availability during a disaster.
Using virtualization for disaster recovery makes perfect sense. Virtual machines (VMs) can be packaged with your operating system, applications and configurations and sent off to a secondary location and take over immediately during a disaster at the primary data center site.
"One of the key advantages is hardware independence and portability so that if you lose the server, you don't lose your app configurations, OS installation, etc. and have to rebuild it all and restore data," said Jon Bock, a senior product marketing manager at VMware Inc. "By backing up VMs with all of that data, you don't have to worry about this stuff, because you are hardware independent," he said.
But setting up a disaster recovery, or DR, environment that promises continuous availability means having top-of-the line virtualization software that many companies can't afford.
One New Hampshire-based VMware user and network engineer who is responsible for building a DR environment said he has had a tough time because the "bean counters" at his company won't let him upgrade to a version of VMware that includes disaster recovery tools such as VMotion or High Availability. So he has to create his DR environment manually.
"Having to rebuild application servers in a limited admin environment typically leads to outside vendors assisting with [time and materials] contracts -- hardly the fastest way to get DR done," the network engineer said. "DR is so difficult to do manually that any form of templating and automation will be the only way all but the largest enterprises get a viable plan."Levels of recovery
At the simplest level, a DR plan gets things back up and running after a failure. There are different levels of availability from which to choose, and the top considerations are an IT shop's service-level agreements, resources and budget constraints.
Unfortunately, the levels of recovery directly correlate to how much money you can dish out, according to Greg Shields, an IT expert and the co-founder of Concentrated Technology, a Denver-based consulting firm.
"There's a general spectrum of cost and RTO [recovery time objective]," Shields said. "By spending a lot of money, you'll get a short (or nonexistent) RTO. By spending not a lot, you can still have DR, but you'll be back to operations in a matter of days instead of seconds."
According to Concentrated Technology, there are four tiers of disaster recovery for virtualized environments, and the better the availability, the higher the cost:
- ($)With the most basic and cheapest disaster recovery plans, your data is restored in days or even weeks. This option is usually satisfactory for a testing and development, but not a production, environment.
- ($$) If you want to be back up and running within a few days, it costs more. This level is OK for internal applications as well as analytic applications.
- ($$$) Immediate availability enables a company to get back up in minutes to hours. This level of disaster recovery capability is typically needed for infrastructure services, support services and messaging services.
- ($$$$) The most expensive option is continuous availability, which offers immediate failover. This level is used for business critical databases and transaction processing appliances.
With the basic, free VMware ESXi, IT can put VMs on a backup server, but backups have to be managed manually. The larger the environment, the more complicated it is to create a DR environment manually.
A step up is the VI Foundation Edition (at $995 per two processors), which gives you VMware consolidated backup to centrally back up VMs. In a larger environment, this capability is useful because you don't have to back up every VM individually and try to manually manage them on your own , said Bock.
"The limitation [of VI Foundation Edition] is scalability, though; if I am starting up 50 VMs in a DR site, I would have to one by one assign VMs to their respective server, because it does not have VMotion. I would probably have to sit down with a spreadsheet to figure this out," he said.
With VMware Infrastructure Enterprise Edition (at $5,750 per two processors), DR is simplest; it includes VMotion and Distributed Resource Scheduler (DRS), which places VMs on the appropriate servers automatically, Bock said.
VMware also sells Site Recovery Manager (SRM), which enables IT shops to automate a DR plan, initiate that plan with a mouse click, and pre-program the sequence in which VMs are brought online at a disaster recovery site without having to write and maintain scripts that automate the DR plan. SRM is not sold as part of a bundle and costs $1,750 per single socket.DR with Citrix XenServer
With Citrix Sytems Inc.'s XenServer, customers can set up a DR environment with the newly free Citrix XenServer Enterprise Edition.
To add high availability and remote DR, IT shops must buy the new Citrix Essentials for XenServer and Hyper-V which costs $1,500-$3,000, depending on the edition.
The XenServer package includes the hypervisor with support for Windows and Linux guests, live migration and resource pooling, and unlimited multiserver management. At VMworld Europe 2009 last week, Citrix's CTO Simon Crosby compared the features in the newly free XenServer to those in VMware Virtual Infrastructure 3 (VI3) Enterprise Edition, which lists at $5,750 per dual-socket system. But to get the high-availability and workload-balancing features of VI3 Enterprise Edition from Citrix, users have to purchase the Essentials Edition.
With the basic version of XenServer for disaster recovery, you need hosts at both locations and storage replication within a storage area network. This functionality calls for products like NetApp SnapMirror and data replication capabilities like those from SteelEye Protection suite. You also need a schedule for metadata backup, which is a free capability of XenServer. With this option, the time it takes to restore the VM metadata depends on the number of VMs.DR with Hyper-V
The network engineer said that because he can't afford fully automated DR with VMware, disaster recovery with Hyper-V was a consideration. But nailing down the options and pricing has proved difficult, he said. Microsoft declined a request to discuss its virtualization DR products and pricing. The company website offers information on continuity and pricing.
Despite Microsoft's reluctance to offer up its DR software details, Shields, who recently wrote a tip about disaster recovery with Hyper-V said that Microsoft's Hyper-V includes native backup features make it a solid platform for disaster recovery.
To create a disaster recovery site with Hyper-V, Shields said you need a backup site, alternate servers and virtual platform software to host failover VMs. You also need virtual backup software, which Hyper-V has via its built-in Volume Shadow Copy Service (VSS) integrations that ensure backup VMs restore successfully, according to Shields.
"VSS ensures a quick restoration of the virtual machine's operating system and applications like SQL and Exchange that reside on top of a VM," Shields said.
Since Hyper-V does not yet support live migration, replication software is necessary so that failover VMs can be moved to the backup site. This can be done through the manual transfer of tapes (or large hard drives), or for more money, you can use the automatic replication of backups over a network for faster failover, Shields said.
There are also several third-party disaster recovery products for server virtualization, including Double-Take Software's Double-Take, which addresses the use of Microsoft Hyper-V for DR, and SunGuard also offers SunGuard Availability Services DR software for VMware virtual servers.
Let us know what you think about the story; email Bridget Botelho, News Writer.