Server virtualization has taken a lot of the pain out of disaster recovery, but for bare-metal workloads, server disaster recovery remains as complicated and expensive as ever.
Now, some data center managers are trying to simplify disaster recovery (DR) using I/O virtualization and orchestration software.
The firm runs production workloads at each of these data centers, but rather than buying and maintaining hardware at dedicated disaster recovery sites, the data centers also act as disaster recovery targets for one another.
The challenge of that approach became apparent last year when the earthquake in Tokyo prompted a failover to the San Francisco office. While successful, the failover wasn’t as fast or granular as the firm would have liked, said Dergham Dergham, infrastructure architect for the firm.
“The hope with PAN Portability is to minimize the time it takes to do a failover and failback,” Dergham said.
PAN Portability is an extension of Egenera’s PAN Manager software that reclaims and repurposes server hardware in response to a DR event. It does so by applying profile information for that server, including information about its hardware, software and networking and storage connections that is stored in an XML file.
The software works across systems managed by Egenera’s PAN Manager, including blade systems from Hewlett-Packard (HP), Dell Inc. and Fujitsu, as well as Egenera’s own BladeFrame platform, and has a starting price of $2,500 per protected blade.
Other converged infrastructure options are available from Cisco Systems Inc., HP and IBM.
Bare-metal DR promised land
The ability to repurpose server hardware efficiently minimizes the number of dedicated servers that must be maintained to take over production in the event of a disaster, said John Humphreys, Egenera vice president of marketing. Because the process is automated, it removes a lot of human error.
But server disaster recovery has room for improvement.
For example, Bingham McCutchen is beta testing an upcoming PAN Portability feature that will allow it to fail over individual server profiles, rather than the entire protected chassis. That feature will give the firm greater leeway about which servers to replicate and protect.
However, the holy grail of server disaster recovery remains out of reach: enabling failover between dissimilar hardware platforms.
In virtualized environments, the problem of dissimilar hardware is largely gone, since the server virtualization layer provides and abstraction layer allows virtual machines (VM) running on one platform to fail over a VMware Inc. host running on unlike hardware.
But the problem is still very real in physical, bare-metal environments where IT teams have to make sure hardware stays in lockstep between the primary and DR sites, Egenera’s Humphreys said.
The problem is particularly acute for companies that grow via acquisition, said Rachel Dines, senior analyst at Forrester Research. “Companies end up with all different kinds of hardware,” she said.
Others say demand for heterogeneous DR software isn’t significant. But the mobility Egenera’s software provides is compelling, said Dines.
“It provides the classic benefits of server virtualization without being server virtualization,” she said.