Storage Area Networks Still on Washington Wish List

by Jon ToigoMilton Halem has a problem.As the assistant director for information sciences and chief information officer at NASA's Goddard Space Flight Center, Halem knows that the data being stored on NASA's computers is growing exponentially. His Mass Storage Team, in fact, just completed a two-year analysis examining the storage needs of NASA's Earth and Space Computing Division, which supports more than 300 research efforts and more than 1,000 users throughout NASA. The study concluded that, by 2004, the division will need the capacity to store more than 130 million files totaling nearly 5 petabytes of data. On a given day, nearly 1.8 terabytes of data will be retrieved from storage for use, while nearly 2.7 terabytes of new data will be added to the massive data pool.By 2004, Halem knows that Goddard will need a storage area network to store and manage the accumulating data. A storage area network, or SAN, can be roughly described as a group of storage devices linked via a high-speed connection that can be accessed by multiple servers. A SAN serves as a "virtual storage pool" with theoretically unlimited scalability, resiliency, manageability and intelligence.When SANs burst onto the scene in 1997, they were hailed as a panacea with the potential to solve virtually every storage-related problem in the corporate computing universe. And while much progress has been made during the past four years, the storage infrastructure described by the visionaries has yet to arrive. And that's why Halem has a problem: Today's SAN solutions do not meet Goddard's rapidly expanding data storage requirements, especially for heterogeneous server and storage environments.NASA needs "more than the capacity improvements represented by a SAN," he said.To address this problem, Halem is preparing to issue an invitation to vendors and integrators within the next six months to join a cooperative venture with NASA to develop the required open, platform-agnostic SAN solution that he needs. NASA in the past has supported the cooperative development of cutting-edge technology, said Halem. In this instance, Goddard will leverage the National Aeronautics and Space Act of 1958, "which basically allows us to enter into a special relationship with vendors," to develop the open SAN technology required by his organization, he said."We want to become a test bed," Halem said.NASA will issue a cooperative agreement notice, offering to share costs with developers. "If the response is such that there is no exchange of dollars, then the agency can create a NASA special relationship with the vendor or vendors," he said.Halem expects to select a partner by early 2001.XXXSPLITXXX-Most industry analysts agree that current-generation SANs fall far short of the visionary storage architectures first described in the mid-1990s. Early SAN architects at Sun Microsystems Inc. of Palo Alto, Calif., and Compaq Computer Corp. of Houston envisioned a data storage infrastructure that provided storage services in a manner analogous to a public electrical utility. Advocates said connecting servers and workstations to a SAN would be as simple as plugging an electrical device into a household wall socket. The open architecture of the SAN and its support for heterogeneous servers and storage devices would enable ready access to the contents and capacity of virtual storage volumes, providing the ultimate data-sharing solution. In addition to unlimited capacity and shared access, SANs were to deliver a new order of intelligence and manageability to storage. They would be application aware, automatically delivering just the right amount and type of storage required by an application as part of the application installation process. From a management standpoint, said advocates, SANs offered numerous advantages over other storage architectures, such as server-attached storage or even network-attached storage (NAS). Unlike other storage topologies, storage devices in a SAN would be centrally located, delivering economies of scale in the resources required for managing storage. Moreover, SAN intelligence would permit disk volumes to "grow" to meet new storage requirements. Simply add more disks to the SAN and the storage volumes available to applications would expand, eliminating capacity planning issues.Finally, SANs would enable greater data protection. Placing storage into its own network, separate from the corporate LAN, would enable the restriction of access to authorized users. Moreover, with a SAN, backups and mirroring operations could occur while production systems were in use and without impacting LAN performance.Current SANs, by contrast, comprise storage networks that are limited in their support for either server or storage subsystem heterogeneity. The explanation, according to most industry observers, is that vendors have jealously guarded their customer base, fielding solutions that work with their branded products exclusively.Added to proprietary architectures are a raft of interoperability and cost of deployment issues deriving from the current "plumbing" of SANs. Most current SANs use the Fibre Channel protocol as the mechanism for data transport between servers and storage devices. Recently completed Fibre Channel standards continue to be implemented differently by different vendors in their products, leading to problems in data exchange between the Fibre Channel SAN switches of competing vendors. Also, Fibre Channel itself is a foreign protocol in most IT organizations, creating a requirement for staff education that adds to high SAN deployment costs. Alternative SANs based on TCP/IP and Infiniband are under development and may solve some plumbing problems.XXXSPLITXXX-The shortcomings of current-generation SANs have not stopped Col. Bill Hose, deputy chief of staff for information management with the Minnesota Army National Guard in St. Paul, Minn., from pursuing a SAN infrastructure. In June 1999, Hose went searching for a replacement for optical storage arrays ? called jukeboxes ? which provided storage of electronic images of personnel records for active and retired personnel. The existing optical storage system was becoming increasingly burdensome to Hose's operation. Data stored on the jukeboxes was difficult to back up, the discs themselves were prone to failure and data loss and the operation of the storage solution was too slow.Hose says that an examination of alternatives led him to consider network attached storage devices from Hewlett-Packard Co. of Palo Alto, Calif., as well as large, multiserver arrays from EMC Corp. of Hopkinton, Mass., Hitachi Data Systems of Santa Clara, Calif., and XIOtech Corp. of Eden Prairie, Minn. The latter group of vendors described their solutions as "SANs in a box" ? large-scale storage arrays featuring Fibre Channel connections between internal disk drives and/or capabilities for linking multiple arrays via Fibre Channel into a storage network. The vendors also emphasized speed of operation and networking capabilities that appealed to Hose."We learned that SANs were faster in operation than NAS devices," Hose reported, "And we like the idea of mirroring [storage volumes] on the fly without impacting performance." Hose says that the appeal of making backups while systems were operating, without affecting application performance, was a key advantage of storage area networking for his environment. In August, using a Department of Interior contract as an acquisition vehicle, the Army National Guard purchased a "SAN-in-a-box" Magnitude array from XIOtech for approximately $80,000. Department personnel, assisted by XIOtech technical support representatives, deployed the 500-gigabyte storage array in Hose's data center in October.Twelve Microsoft Windows NT Servers are connected to the array via point-to-point Fibre Channel connections, said Hose. The balance of his servers ? Unix servers in particular ? continue to have their own storage subsystems.Today, the XIOtech SAN serves up data to applications used by approximately 160 users to administer such functions as soldier payroll, promotion evaluation, activation, training and logistics for more than 10,000 personnel. Hose said he was pleased with the array, and enhanced its capacity to a terabyte of storage in February 2000, while upgrading point-to-point server connections to switched connections in June of this year."We deployed a Gadzoox Networks [of San Jose, Calif.,] SAN switch to connect a tape backup system with the servers and XIOtech array," Hose said, noting that the new configuration enabled him to achieve his real-time data backup objectives. He said that XIOtech's Redi Copy backup software, as well as the array's management software, are "intuitive and easy to use."The current SAN will handle his storage requirements for "the next three to four years," Hose said. Additionally, he is considering the purchase of a second half-terabyte XIOtech "SAN-in-a-Box" for use at Camp Riley, near St. Cloud, Minn.Hose acknowledged that he was unsure whether the storage array from XIOtech could co-exist with other disk storage arrays in a SAN, nor was he sure whether data from his Unix systems would ever be included ? or could be supported ? in the XIOtech SAN. Asked whether he was concerned that the XIOtech solution wedded the Minnesota Army National Guard to a single vendor for future storage products, he said that any concerns he may have had were dispelled by the company's acquisition in January 2000 by disk drive maker Seagate Technology Inc. of Scotts Valley, Calif."Now that Seagate bought them, we're comfortable that they will continue to grow their product line and comply with standards," he said.XXXSPLITXXX-Like Hose, John Marshall was also drawn to SANs as a means to meet data backup requirements. The enterprise systems department head at the Office of Naval Intelligence National Maritime Intelligence Center in Suitland, Md., Marshall said that his turnkey SAN, purchased for approximately $500,000 in April from integrator Marzik Inc. of Lanham, Md., will go online in October, providing shared backup for ONI's Unix-based Oracle database servers.ONI's Fibre Channel SAN is the latest addition in an ongoing infrastructure improvement effort that dates back to 1995, according to Marshall. He said that the original objective of the effort was to develop "a network-centric approach for hosting our applications [and] to share access between Unix and NT systems to files and applications." Marshall said that the new infrastructure, including servers and storage devices, "was totally engineered in-house, but implemented using contractors."In March 1997, ONI deployed network-attached storage systems from Network Appliance Inc. of Sunnyvale, Calif., to facilitate Unix/NT user file system sharing, Marshall recalled. Unlike a backend SAN, NAS storage installs directly onto ONI's TCP/IP local area network and uses a network file system protocol such as the Network File System (NFS) or Microsoft's Common Internet File System (CIFS) to provide access to file volumes to servers and workstations. Since the initial deployment of "a couple of hundred gigabytes" of NAS storage, Marshall said, ONI has added new Network Appliance NAS filers providing more than a terabyte-and-a-half of NAS-based storage capacity.He observed that ONI's Unix-based database systems have different data access requirements and are equipped with their own internal and external storage subsystems, rather than using NAS storage. This difference created a concern about data integrity.He said that the Network Appliance filers, which were acquired under the General Services Administration schedule pricing at a cost of approximately $1.5 million, provided their own internal backup using disk mirroring. Unix database systems storage, by contrast, was not mirrored. "We needed to perform backups to a tape drive on each server separately," Marshall explained, emphasizing that this laborious and time-consuming process required that the servers and applications be quieted "for several hours to back up roughly a terabyte of data." Marshall sought a means to centralize the management of backups for the Unix systems and settled on a SAN. ONI's new SAN-based backup solution for the Unix servers and databases uses Veritas Software Corp.'s NetBackup for Unix software to copy data to a shared tape library system from ATL Products Inc., a subsidiary of Quantum Corp. of Milpitas, Calif. The tape subsystem may also be leveraged to back up some data from the Network Appliance filers via a direct connection.Marshall noted that the Unix backup solution has increased his interest in exploring the potential of SANs in other areas, but he is not prepared to abandon his reliable NAS architecture just yet. "Network Appliance is talking about a new protocol, Direct Access File System, that might provide the means to add our Unix databases to our NAS infrastructure. Alternatively, SAN technology is evolving and may eventually provide a stable architecture for [re-hosting our applications and file systems data.] We are taking it real slow. We will wait and see which way the technology goes. We are currently very happy with our mixed storage infrastructure."XXXSPLITXXX-It is exactly a mixed storage and server infrastructure that creates stumbling blocks for SANs at Goddard Space Flight Center. Halem said he is confident that a joint project between NASA and the vendor community could produce a truly open SAN infrastructure within two to three years. Necessity, in part, forces Halem to be optimistic about the results of the initiative."We need to develop three SANs that will serve Space Science Directorate, Earth Science Directorate and Flight Projects Office, possibly with [mutual] backup capability," Halem said. "At some future point, we will integrate the little SANs into a big SAN. All of these organizations would benefit from the capability to share data with each other. Instead of physically moving and networking data, its centralization in a SAN with common access would provide tremendous [cost- and time-savings] advantages."The open SAN technology, when developed and tested, would belong to the development partners, he said. In addition to providing a data-sharing solution for the directorates and offices at Goddard Space Flight Center, he hopes to see the new open SAN technology applied to NASA's network of Distributed Active Archive Centers. Halem concedes that some of NASA's storage requirements ? including a requirement to archive climatological data for access over a period of decades, or even centuries ? may not be applicable to a business environment, where Internal Revenue Service mandates require a mere seven to 10 years of accounting data storage. Nevertheless, he believes the technology will find a home in any organization where investments have been made in a broad range of storage and server products.In the business world, as in NASA, to paraphrase Halem, the ability to collect data is growing faster than the capability to store it. To Halem, open SANs are an important component of any IT strategy seeking to keep pace with change.