The falling cost of high availability

Businesses are demanding faster recovery times and finer recovery points, the number of critical applications they need to protect is on the rise, and the volume of data that must be backed up is exploding. And yet, by almost any measure, the cost of high availability and business continuity is falling.

Enterprise-grade storage has gotten significantly cheaper, and so has bandwidth. Even the software required for disaster recovery costs significantly less than it did just a few years ago.

“Previously just to protect maybe three terabytes to ten terabytes of data cost over a million dollars in software, and today you can do it for 30 to 50 thousand,” says David Self, vice president of systems engineering at InMage Systems. “Prices are definitely coming down from a bandwidth, storage, and software standpoint.”

Self points to a fourth factor that’s helping to drive down the cost of high availability: server virtualization. He says that roughly 75 percent of InMage Systems customers are using virtual servers in their disaster recovery sites, leveraging the support of InMage’s business continuity solution for physical-to-virtual fail-over.

Vendors such as PlateSpin and Scalent Systems are also helping customers reduce the number of physical servers required for disaster recovery. At the same time, server virtualization platforms — most notably VMware — are helping companies such as PTC achieve high availability within the walls of the datacenter.

A software company in Needham, Mass., PTC’s business continuity strategy involves creating standard configurations, managed using Opsware tools, and ensuring that virtual images of those configs are all catalogued and able to be recreated wherever they might need to be.

“We standardized on VMware for both Linux and Windows using VMotion across the board,” notes the company’s vice president of IT, Greg Wolf. “Everything has a hotsite within the same datacenter.”

For providers of disaster recovery and hotsite services such as SunGard Availability Services, virtualization technologies provide ways to offer customers more flexibility at lower cost. John Lindeman, vice president of advanced recovery product management at SunGard Availability Services, says the company is tapping virtualization in systems, storage, and the network to fit customers more quickly and readily into SunGard’s fail-over and recovery environment.

“The whole discipline of adopting [customers’] production to our recovery has been an imperfect science because we may not have the exact components as they have in the flavor they have it,” Lindeman says. “With virtualization we can adopt more quickly and more cost effectively.”

SunGard has been using virtualization in its data networks and storage networks to weave together customers’ infrastructure and its recovery platforms, “so it looks to the customer as one big, ubiquitous network,” Lindeman says. “At the storage level,” he points out, “we’re seeing some real big enhancers especially in the mid-tier area of virtualization, where we can go from one topology to another — from a Hitachi to an EMC, from an IBM to a Hitachi — and be able to utilize our assets better. We can look at carving up physical storage across multiple customers for fail-over and recovery use. We’re not quite there yet, but we see [storage virtualization] as a vehicle to get us there.”

As part of SunGard Availability Services’ recently announced virtualization and automation initiative, the company has begun supporting VMware ESX as one of the enterprise-class operating systems it manages for customers. According to Don Norbeck, director of product development at SunGard, the ESX support will be released as a productized offering within six months, and the company is working to develop virtual-to-virtual recovery services in the same timeframe. 

“For customers that want to bring up a virtualized environment on a traditional hotsite, that requires a number of additional steps,” Norbeck says. “We have the process in place to support those steps, and more importantly the staff there that can support them and turn those up more quickly than waiting for a customer to arrive at our doors.”

Norbeck says that the first phase of the initiative is to provide assisted recovery involving dedicated services to individual customers, where SunGard would be doing about 30 percent of the work and the customer about 70 percent. The next phase, a managed recovery services offering projected for sometime next year, aims to reverse that workload. For the managed recovery services, SunGard will do the balance of the work, and it will also select a set of standard components for the recovery platform — the storage, infrastructure, monitoring, and management technologies, as well as the virtualization platform, converters, and players.

Finally, it’s worth noting that physical-to-virtual and virtual-to-virtual fail-over aren’t the only ways that server virtualization can help drive down the cost of disaster recovery. One international financial organization, where Tom Ferris serves as manager of servers and storage, is a case in point.

“Virtualization is big in our overall disaster recovery strategy from the perspective that in the business continuity center we have a lot of the core services running on virtual machines,” Ferris says. “We don’t have separate physical servers for every domain controller, every DNS server, every authentication server. Those are all running on virtual machines, and we’ve been able to reduce the number of physical servers that we need quite a bit.”