10/9/09 By: Christian Perry
A segment of storage in almost every data center skirts by every day without doing much work. But through the use of partitioning, it’s possible to get that storage back to work and keep it there.
“Properly partitioned hard disks will allow the data center to maximize its storage investments by reallocating unused disk space and consolidating data, resulting in the need to purchase less new storage,” says Jim Thomas, technical services manager for Paragon Software Group (www.paragon-software.com). “Increased system performance can also be noticed through defragmentation of partition contents and the MFT [Master File Table].”
• Partitioning can help data centers deploy previously unused storage space for applications, testing, and other tasks by dividing hard drives into separate storage areas.
• Although the actual partitioning process is simple, experts recommend planning before conducting partitioning sessions to determine the best use for the technology and prepare for potential changes.
• Partitioning can force drive letter assignment changes, conflict with existing file system problems, and cause other issues, so data center personnel should expect the possibility of some problems with the technology.
At its core, partitioning is the process of dividing hard drives into separate storage areas, or partitions, to make use of previously unused disk space. According to Curtis Breville, data storage evangelist for Crossroads Systems (www.crossroads.com), partitioning was originally designed to dedicate part of a disk drive to a specific purpose to allow the data to be physically close together and speed up access to data on a device that used random-access searching.
“Partitioning also allowed for better use of disk space and kept one application from taking away space needed by another. With astute planning and accurate growth prediction, each application would have the right amount of storage, and there would be less wasted disk [space],” Breville says.
Today’s flexible partitioning technologies continue to build on that performance-enhancing tradition, delivering automated and unattended operations, RAID support, dynamic disk support, Windows-based tools for on-the-fly partitioning, and even bootable recovery media to enable partitioning operations while systems are offline. Also relatively new is thin provisioning, which allows partitioning without the need to physically allocate storage at initial setup.
Data center managers who neglect to implement partitioning for fear of disrupting delicate system environments might be pleased to learn that partitioning can occur while systems are online. However, before moving ahead with partitioning, experts recommend some basic planning procedures to ensure that the technology is working to its full potential.
“Typically, after the goals and business case for partitioning have been established, history performance data on existing servers and applications is collected to assist in the planning process as well as information on the importance of each application to the business,” explains Gary Thome, director of strategy and architecture for Infrastructure Software and Blades at HP (www.hp.com). “Architectures and partitioning software are chosen based on the goals of the project, along with plans for management, high availability and disaster recovery, and backup and security procedures.”
Thome also recommends determining the metrics the data center uses (or will use) to measure success. For example, is IT judged based on response time to end users? On percent of unplanned downtime? On costs of capital expenditures or of the power bill? Also, data centers planning to implement partitioning should gather utilization data from their existing servers, storage, and applications, Thome says.
The actual process of partitioning new or existing drives is surprisingly simple. “Most partitioning utilities show each hard drive in the system with graphic representation of the partition layout. After installing the partitioning software, an operation such as resizing partitions is usually as easy as dragging the border of a partition to the desired size or entering the desired size of the partition in numerical form, before allowing the application to carry out the partitioning operations behind the scenes,” Paragon’s Thomas says.
Rolling partitioning into production—that is, moving programs and data into a partitioned environment—can be accomplished with tools that automate the transfer of applications from physical servers to virtual servers, Thome says. From there, data centers can use ongoing monitoring and capacity planning to ensure the optimal distribution of workload and resources.
Although partitioning is generally a safe process, it’s not without pitfalls. For example, Thomas warns that when booting a server from recovery media, drive letter assignments might display differently than how they appeared in the host operating system. Further, he warns that file system errors and bad sectors can cause numerous problems, so it’s wise to check for physical errors and file system errors before creating or modifying partitions.
James Wilson, product manager for HP StorageWorks, says that another concern with storage cache partitioning is that the time required to move cache is variable and does not address short-term hot spots or sudden changes in workload. Further, the cache being moved is not available to any partition from the start of the move until the cache is reassigned to the new partition.
Despite these potential drawbacks, partitioning is here to stay in data center environments as an effective method for increasing operational efficiency. “Partitioning is like cutting a child’s birthday cake,” Thome says. “As long as you plan ahead and measure carefully, everybody is going to be happy.”