Virtualization technology is changing the face of disaster recovery (DR). Traditional DR-based backup and restoration techniques typically imposed high IT costs as administrators anguished over delicate hardware dependencies -- often forcing some companies to ignore or overstate their DR preparedness.
Virtualization itself adds complexity to the environment. But it also adds consolidation and flexibility, allowing organizations to copy live storage and server workloads between remote sites running different hardware.
Backing up a workload and restoring it to another machine may seem like a simple matter, but myriad practical concerns further complicate this objective. Central issues include disruption, hardware dependencies and the cost implications of both.
Migrating copies of workloads between physical hardware is disruptive, limits application and database availability and can potentially result in lost productivity or sales. In addition, workloads and storage volumes are notoriously sensitive to physical hardware. The slightest difference could easily cause operational problems with workloads at the warm/hot remote site as well as with restoring them to the primary site.
Errors can dramatically extend recovery beyond the anticipated recovery time objective and stress out already overworked IT staff. Typically, companies would absorb the cost of duplicate hardware and upgrade expenses. Some organizations have tried to mitigate this cost by omitting noncritical workloads from the DR plan, relying instead on the availability of common backups for later restoration.
Benefits of using virtualized servers for DR
There are several benefits to virtualization; hardware independence tops the list. Virtualization creates a layer of software abstraction -- the virtual operating system or hypervisor -- between the workload and its underlying hardware. This decouples the two and allows workloads to reside or operate in a virtual instance unrelated to the server and storage hardware underneath.
This enables a physical or virtual server working in the primary data center to be copied into a virtual instance on almost any other server hardware located at the disaster recovery site. This site can run the virtual instance without concern for differences in the remote server's hardware. It then can be restored to the original physical server or migrated to another virtual server.
Such flexibility enhances server availability. When a fault occurs on one virtual server, workloads can be migrated to another available server. Server virtualization also enables hardware consolidation, allowing multiple virtual instances to operate on a properly configured server.
A company can use different servers in the DR site and can operate with fewer physical servers if each physical instance is running multiple virtual machines (VMs). The benefit of reduced hardware becomes more obvious when the primary site is also virtualized. There will be fewer servers at the primary site and virtual workloads can be moved nondisruptively. Synchronization can occur between sites without affecting application and data availability.
Storage virtualization can be included in a virtual DR environment, but it's not a requirement for Ray Luchessi, founder of Silverton Consulting Inc., an independent technology consultant in Broomfield, Colo. Luchessi says that many enterprise-class network-attached storage (NAS) and Fibre Channel (FC) storage subsystems natively support data replication.
Storage virtualization improves flexibility in the creation, control and nondisruptive migration of storage volumes, but it doesn't offer the consolidation found in server virtualization. One terabyte of storage at the primary site still requires one terabyte of storage at the DR site.
The collective benefits of virtualization can save organizations money. There are fewer physical systems to contend with, reduced server hardware costs, reduced switch port counts, and less wiring and floor space is required. A virtual environment also requires less power and cooling. Virtualization allows you to reallocate older servers or storage systems from a technology refresh at the primary site to the DR site, saving additional hardware capital expense.
Virtualization's hardware-agnostic and flexible nature eliminates most recovery problems and aids availability, minimizing downtime and the burden on IT staff. Hardware savings often balance the cost to license and maintain virtualization software.
Drawbacks of virtual disaster recovery
The abstraction layer needed for virtualization will add some complexity to the environment. Because it can be difficult to track virtual problems to a physical source, it requires more management and can complicate troubleshooting.
Consolidation is an even bigger issue. A fault on one physical server will affect all VMs running on that server. It may be necessary to implement high-availability (HA) architectures such as server clustering to guard against hardware faults.
Servers hosting multiple VMs must be configured with CPU, RAM, I/O and network resources adequate enough to support expected utilization. Otherwise, the all - VM performance on the server will suffer. However, this may not be a serious concern for a DR site that's only used for restoration. "They [administrators] could potentially get by with less processing power at the secondary site and [lower] operational costs," Luchessi said. "But they spend a bit more on [virtualization] software to support that."
Stephen J. Bigelow, Senior Technology Writer, can be contacted at firstname.lastname@example.org.