Fifty years after IBM invented the disk drive, storage provisioning
Continue Reading This Article
Enjoy this article as well as all of our content, including E-Guides, news, tips and more.
Storage provisioning is the process of logically carving up physical disk space to meet an organization's need for storage capacity, performance, security and efficiency. It encompasses the assignment of servers/hosts to appropriate storage, specifying network paths between hosts and storage, and masking servers and zoning the network to ensure access by the right servers. The entire task involves dozens of individual steps.
When done correctly, provisioning results in storage that performs well and scales nondisruptively. When provisioning isn't done correctly, application performance can degrade, security is compromised, scalability is inhibited and capacity utilization is low, resulting in wasted storage space (see "Storage provisioning pitfalls" sidebar below).
Complicating the provisioning challenge are today's widely heterogeneous enterprise storage environments: more storage, additional arrays, different types of array devices, multiple storage tiers, more RAID levels, increasing complexity, and many best-practice provisioning guidelines or standards. "Some day there will have to be a standard way to deal with storage provisioning, but there isn't now," says Rhoda Phillips, research manager, storage software at IDC, Framingham, Mass. And, she adds, that day won't be coming anytime soon.
Without a set of best practices for provisioning, storage managers are left trying to hammer out what works best for their organization based on application requirements and the organization's particular mix of storage technology. For most, it's a manual trial-and-error process in which the lessons learned the last time may no longer apply when a new array, large-capacity disk drive or new tier of storage is thrown into the mix.
To get a better handle on storage provisioning, try to figure out who actually takes charge of provisioning. "With provisioning, there are a lot of handoffs," says Robert Stevenson, managing director, storage practice at TheInfoPro, a New York-based research company. Stevenson is a former storage manager at Nielsen Media Research, where he supervised the provisioning of 1,000 LUNs.
At Nielsen, Stevenson would begin the provisioning process by huddling with the DBA whose system needed the storage. Before the job was done, however, provisioning would involve system and network administrators in addition to the storage team.
The DBA knew the performance and capacity requirements, while the system administrator would make sure the hosts could find the newly provisioned storage. The network team handled the zoning, while the storage team did the actual carving up of the capacity. Even within the storage team, different groups might have a hand in the process. "The engineers would carve up the disk arrays. The operations group would do the LUN mapping and masking," explains Dan Weinstein, senior consultant at storage research and consulting firm GlassHouse Technologies Inc., Framingham, Mass.
Effective provisioning requires a coordinated effort (see "Provisioning best practice" sidebar) with good communication among all team members. In smaller organizations, "the storage team or maybe just one person would do it all by hand: allocate the storage in the array, configure the connections and point the hosts at the storage," says Jason Anderson, storage architect at Datalink Corp., a storage integrator in Chanhassen, Minn. Even then, you still need information about the application and its performance, capacity and protection requirements, which will involve input from other people.
"The first thing you need to do before you start provisioning the storage is to understand what is needed," says Weinstein. That means defining the application's service-level requirements in terms of storage capacity, protection, performance and resource management. Without knowing this information, administrators can't make the provisioning decisions that will be required almost immediately.
The requirements must then be translated into storage terms: what kind of storage array and disk, which tier of storage and what RAID level. "Once you determine this information, you can look for appropriate [physical] space before you start carving it up logically," says Weinstein.
Stevenson would kick off the provisioning process by sitting down with the storage administrator and the DBA. Using a standard spreadsheet, they'd translate the application capacity and service requirements into a list of possible LUNs and their associated I/O performance. "This is a lot easier for existing applications that need more storage and have a history of what they have used before," says TheInfoPro's Stevenson. With a new application, the team would make their best estimates based on similar applications. The information on the spreadsheet would then become their guide when they began provisioning storage.
You should also "know the criteria and process by which requests for storage get approved in the first place," says Tim Arland, principal consultant at independent storage integrator Forsythe Solutions Group. Without firm criteria and an orderly process, storage provisioning quickly descends into chaos, with LUNs being carved out willy-nilly any place there seems to be physical capacity and without regard for the performance characteristics of the storage system or storage service requirements of the applications.
Click here to continue reading The right way to provision storage, page 2.