|
 |
|
General Information
|
Add To My Personal Library |
March 13, 2009
• Vol.31 Issue 10
Page(s) 8 in print issue
|
Data Centers On The Move
When Moving Or Reorganizing Your Data Center, Preparation Is Key To Avoiding Problems
|
| Key Points
• Careful planning and having a detailed overview of how the move will proceed will help you avoid problems.
• Make sure you have reliable backups so that you can failover to that backup or restore lost files in the event of a complication and steer clear of having a single point of failure.
• Rank the importance of your applications to avoid serious downtime with the mission-critical ones and to preserve application performance. | | You can’t possibly anticipate every potential problem when you move or reorganize your data center. Swastik Lahiri, principal at technology management solutions provider and consultancy Technisource (www.technisource.com), remembers a time when Technisource was moving a data center to a different location in New York City, and he and his colleagues discovered that the destination data center used organized labor.
“They worked a certain amount of hours, and we had to wait in line behind whoever else was in queue and didn’t have control over how things were being put in,” Lahiri says. “It ended up messing up our schedule significantly. Unless you’ve had it come up, you wouldn’t even have had [that complication] on your radar.”
 Nevertheless, you can avoid many more obvious difficulties through awareness and common sense. Here are some suggestions for how to avoid or resolve some of the more typical problems you may face.
 Test Your Backups Before You Move
Lahiri says that before you move or reorganize, you should make sure you have proper backups of your system so that you are protected in case something goes wrong during the process.
Lahiri also recommends testing your backups even to check for such low-level issues as duplex mismatches. “If the network isn’t configured right for duplex, you’ll get really slow performance, and it’s not easy to identify that problem ahead of time” without testing, Lahiri says.
Avoid A Single Point Of Failure In Your Updated Network
Server and storage virtualization is one of the primary drivers for reorganizing your data center, but you want to make sure that once you have made the move or consolidation, you haven’t inadvertently caused a single point of failure, explains Brace Rennels, technical marketing manager at data protection solutions provider Double-Take Software (www.doubletake.com).
“When physical server workloads are consolidated to a smaller, more efficient virtual host, there is an issue of all of the virtual guests residing on the same physical server, as well as sharing the same shared storage,” says Rennels.
For his part, Rennels recommends finding a software solution that enables you to provide live migrations of entire virtual host workloads to keep your systems highly available. “The solution should be flexible enough to be able to easily move those workloads across greater distances for disaster recovery in the event of sustained disaster at the primary data center,” he says.
Make Sure Infrastructure Requirements Are Met
Virtualization may reduce your overall data center footprint, but don’t assume it will automatically lessen your power and cooling requirements, says Rennels.
“Consolidating to a blade and virtual infrastructure can actually increase heat in a condensed location and cause more cooling consumption in order to reduce those workloads to an optimal operating temperature,” he explains, adding that careful planning is needed to make sure you maximize your power and cooling efficiencies.
Cables must be paid attention to, as well. According to Technisource’s Lahiri, old cabling too often gets crimped and mangled during a move or reorganization, causing easily avoidable hassles during installations. “We just ask for new cabling whenever we do a move and throw away old cabling because it’s not worth the risk,” he says.
Establish A Provisioning Process To Deploy New Workloads
If you are looking to move to more efficient blade technology, then you will want to have a good provisioning process for rapidly deploying new workloads, Rennels also notes. “Installing the OS disks from scratch or restoring from tape won’t be sufficient,” he says. “Using a netboot solution where servers can be pointed to disk images residing on an iSCSI device can boot up the servers immediately.”
Rennels adds that this strategy can also be beneficial from a security standpoint by helping to further protect from viruses or data theft. “This will help make the new data center architecture flexible so that many servers and workstations can be provisioned efficiently and rapidly,” he says. 
by Robyn Weisman
Top Problem: Unexpected Application Performance Hits
As virtualization increasingly becomes the preferred mechanism for relocating or optimizing a redeployment, you need to be aware of potential application performance hits, says Shane O’Donnell, vice president of engineering at dynamic data center solutions provider BlueStripe Software (www.bluestripe.com).
When IT moves a server to another location or makes the conversion from physical to virtual machines, application performance often is collateral damage for that change, O’Donnell says. Recently, one of O’Donnell’s customers discovered through software designed to delineate a system’s performance profile—including application behaviors, dependencies, and susceptibility to new latencies—that one of its primary applications had taken a performance hit because it still had links back to its development systems.
“We advocate with some bias that you actually need to understand the application itself before you can really move it around and expect similar behaviors,” O’Donnell says.
In addition, Swastik Lahiri, principal at technology management solutions provider and consultancy Technisource (www.technisource.com), recommends that you do as much planning as possible to identify what your most critical applications are. “Every department is going to tell you that their stuff is the most important, but you need to get them to force-rank it and tell you these are the most important systems,” says Lahiri. “If there’s something that just can’t be down, you’ll need to come up with a plan [where you] almost failover to the backup while you move to the primary system and then switch it back,” he says. |
|
|
