Spring cleaning for storage clutter

Information technology executives may be tempted to keep buying more storage devices as online data continues to accumulate at alarming rates. However, adding new capacity to servers generating the most storage demand won’t necessarily solve the problem.The use of available space on server-attached storage is notoriously poor. And the need to manage myriad storage systems only adds to the expense and inefficiency of the storage infrastructure.Storage consolidation, which can be done by using fewer but larger storage devices or centralizing data in a storage-area network (SAN), provides an alternative strategy that is putting many government agencies on the path of cost avoidance. Clearly something must be done. And it’s a foregone conclusion that storage costs will continue to increase with the growth of data, said Rich Harsell, regional director at GlassHouse Technologies’ Federal Division. But with consolidation, an enterprise has an opportunity to at least keep the spending in check. Harsell estimated that a large organization might be able to scale back future storage costs by as much as 30 percent through consolidation.“There’s always a positive slope,” Harsell said of storage costs. “It’s the control of the slope that is the important aspect.”But projects must be carefully planned — and pitfalls avoided — for organizations to reap the benefits of consolidation. Here are the steps that experts recommend to make a consolidation project go smoothly.Industry consultants advise that a storage consolidation project should always start with a thorough understanding of the existing environment. It’s difficult to plan for the future when the present is not well understood.“Typically, one of the biggest hurdles is that [government agencies] have no idea of what they have,” said Larry Fondacaro, senior solutions architect at integrator Emtec Federal.That situation can be rectified by an analysis of server-attached storage. Armed with the knowledge of how much storage is in each server, agencies can determine which servers are good candidates for consolidation, said Howard Weiss, field solutions team manager at CDW. For example, servers with large databases or file servers with considerable back-end storage would be candidates for consolidation. A small application server that stores nothing beyond its operating system doesn’t belong on a consolidated SAN, Weiss said. “You don’t want to put every server on a SAN,” he said. Meanwhile, hardware inventories also lead to a better understanding of utilization rates, which is helpful for setting future goals. For example, officials at the University of New Hampshire started a consolidation project last year to corral storage on a single SAN.Before beginning the project, the school discovered that its average storage utilization in its direct-attached environment was about 40 percent, said Joe Doucet, director of the Enterprise Computing Group at the university’s Computing and Information Services department. Utilization was also inconsistent; some server-attached storage was at maximum capacity, while other devices used only 30 percent of the available space.Now, the university aims for a 70 percent utilization level, an objective it plans to reach by the end of 2008, Doucet said.Organizations must know the performance requirements of their applications and user expectations. Storage performance is measured in I/Os per second, the amount of data transferred on and off a hard drive.An organization may be managing by simply by relying on the speed of the five drives on a typical server, in which each drive is capable of 100 to 150 I/Os per second. Those rates are much lower than most low-end SANs, Weiss said. “A lot of customers buy the faster SAN because they think they need it,” he said. The lesson is don’t overpay for performance you don’t need, unless you can make a solid case for needing the extra capacity in the future. An agency that determines its current storage costs can calculate the benefits of consolidation and build a better business case. Knowledge of the cost baseline is also handy for evaluating alternative storage solutions. However, agencies often lack this knowledge because it’s hard to collect.“The organizations we’ve dealt with don’t have any cost models in place, and hence the only costs that they can put a finger on are acquisition costs,” Harsell said.Meanwhile, the initial purchase price of storage provides only part of the picture. Harsell estimates that 20 percent to 30 percent of a typical storage budget is tied to acquisition. Other significant costs include salaries, data center floor space, power and cooling, and software license fees.“Frequently, the comparison between what we are getting rid of and what we are bringing into the data center is pretty narrowly looked at in terms of total cost of ownership,” said Bob Wambach, senior director of storage product marketing at EMC.Wambach ranked labor and utilities as the top two storage-cost items. Acquisition, over the life of a storage solution, can end up as the No. 3 cost source, he said.Agencies sometimes overlook the storage-related work server and networking employees do, and they may fail to include their work in calculating overall storage costs, Harsell said. By getting a detailed grasp of cost, a storage shop can become an in-house service provider if it chooses and can charge other departments for the storage capacity they use. Harsell said GlassHouse encourages its clients to adopt Information Technology Infrastructure Library-compliant cost models. ITIL is a set of best practices for managing IT services. It’s not enough to account for server-based storage. Agencies need to understand the nature of the data they store. “Most people I talk to really don’t know the data that is in storage on the servers,” Weiss said.The result is inefficient and expensive storage practices. For example, an organization that doesn’t identify static data — for example, JPEGs or PDFs — stands to make thousands of copies of unchanging files during years of back-up sessions, Weiss said. But an organization that flags static data can design a SAN with an area for fixed-content storage, he said. Agencies that take the time to sort their data can usually take advantage of tiered-storage architectures. In such arrangements, technicians assign data to the most cost-effective storage platform based on the data’s criticality, or in the case of fixed-content storage, its changeability.Critical data goes on the top tier, which is typically the highest performing disk storage. Noncritical data may reside in the bottom tier, a tape — or disk-based archive. Some organizations operate an intermediate tier for data of middling value. This tier usually consists of storage built around lower-cost devices, such as Serial ATA (SATA) disk arrays.“Normally, we try to keep the high-performance applications like [Microsoft] Exchange…on Fibre Channel, and depending on the performance expectation, that could be a 10,000 or 15,000 rpm drive,” said Dale Wickhizer, chief technology officer at Network Appliance’s federal division. “For most everything else…SATA is a great candidate.”Fondacaro said Emtec Federal uses storage resource management tools to classify data for customers. Data may be classified by application, file type and most recent access time. That arrangement helps customers decide whether a particular piece of data is critical and needs to be housed in high-end production storage or whether it is infrequently accessed and can be housed in near-line or archival storage. An organization that understands its storage environment, costs and data types can proceed to the next step: evaluating vendors and deploying a storage solution. At this point, no single approach is indicated. Weiss said options range from “small, budget-conscious products all the way up to enterprise SANs.”Lt. Col. C. J. Wallington, division chief of advanced technologies at the Army’s Program Executive Office for Enterprise Information Systems, said he has identified a storage consolidation need for small groups of users. The division is testing an EqualLogic SAN for tactical use. On the opposite end of the scale, SANs that already provide some degree of consolidation are evolving into more dense configurations, Wambach said. The same holds true for network-attached storage devices, he said. Those devices consolidate file-level storage, while SANs aggregate block-level storage associated with databases. “The consolidation trend is to build bigger boxes,” Wambach said.