One consequence of running vast computer networks is the greater energy needed to keep systems operating and cooled. Recognizing that computing capacity and power are wasted when a data center runs multiple servers, technology companies are re-configuring data centers so that they run cooler and simplify disaster recovery.
California electric companies today offer substantial rebates to firms that reduce their number of servers. This is possible as enterprise software leveraging storage area networks allows servers to be deployed as files that can be distributed across a of pool of servers. Because a file need not be housed in a single server and is now decoupled from the machine, far fewer servers are needed to run multiple programs.
Smart Business spoke with Omar Yakar of Agile360, a technology-consulting firm, about the impact the development will have on data centers.
How did ‘virtualization’ of servers come about?
The concept has been around for maybe 20 years, primarily in Unix and mainframe applications. But for four years now industry has software that will run on X86 servers, the ones that run Windows and Linux programs. One company in particular, VMware Inc., has taken a leadership position.
The challenge businesses face is to manage its increasing appetite for information technology cost-effectively and, just as importantly, to quickly and effectively replicate not only data, but entire systems during disaster recovery. In our world today, this has meant a new server every time we need a new application. What has resulted is that data centers now are chock full of servers, which often are under-utilized, but needed nonetheless, to support numerous functionalities. It is not unusual to see that the addition of one new software application results in the addition of many servers to support the program. The consequence is crowded server racks; vastly higher energy costs from running so many machines and from cooling increasingly heated data centers; and the vastly complicated data replication efforts that result when disaster strikes.
How does virtualized software help retrieve data and reduce energy costs?
Although it comes as a surprise to many IT managers, electric companies report that an average data center today consumes as much electricity as 700 houses. Adding to the problem, the cost to cool a server is estimated to consume 1.2 to 1.3 times as much electricity as the server itself.
In a virtualized word, what used to be known as servers are now logical representations of servers in the form of files and sit on a storage area network, allowing software, such as VMware, to sever a file’s wedded tie to a single server, thereby, allowing a file to be accessed from any server within a data center. Because a file can now be distributed across a pool of servers, fewer physical servers will be needed to support a broad range of computing needs. Our experience shows that a company can achieve a 10- or 15-to-1 consolidation ratio through the use of virtualization software, and in the process, reduce power consumption to run and cool a data center while decreasing the strain on power plants.
PG&E offers rebates to data centers that reconfigure their servers by using virtualization software, and SDG&E soon will. This is on top of the $300 to $600 in annual energy costs that the electric company expects a customer to save for each server that is removed. Those savings can more than double when reduced data center cooling costs are also taken into account.
The potential for savings in capital equipment is huge. In one study we undertook for an area hospital, we concluded that the facility’s 117 servers could easily be pared down to 14. With fewer servers running, the replacement costs of those servers will be cut to a fraction of what it now is, because not only are fewer units needed at a primary location, the size of the secondary back-up facility also shrinks. The gains to industry are enormous.
How can most companies achieve the benefits of such a scenario?
You need someone who can implement the software, know how to apply for the electric company rebates, redesign the architecture of a data center and possibly how the procurement of servers is now cost allocated, select the kind of servers needed, figure out the many finer points of reconfiguring a network, and, finally, train the in-house IT staff.
With virtualization software, disaster recovery grows exponentially faster and massively cheaper. It becomes reasonable even for the little guy to operate a back-up center. And, significantly, we calculate that the provisioning process of setting up a data center also can be dramatically reduced. That happens once we study a data center’s operations and map out a new plan that consolidates computing operations to run on the fewest possible units. The result: a four to six week process can now be completed in as little as four hours.
OMAR YAKAR is the co-founder and CEO of Irvine- headquartered Agile360. Reach him at (949) 253-4106 or firstname.lastname@example.org.