January 26, 2004 7:30 AM PST
Data storage worlds uniting
EMC, IBM and less well-known storage specialists such as BlueArc are beginning to mix and match the best elements of network-attached storage (NAS) devices and storage area networks (SAN) to improve the efficiency and lower the cost of storing data.
NAS boxes, which hook into existing Ethernet networks, are simpler to manage than SANs, but provide less performance and can top out on capacity. SANs generally are far faster and hold more data, but they can be expensive and result in cumbersome file sharing.
Companies are embracing products that marry so-called SAN and NAS storage technologies in order to squeeze more out of existing gear and lower their management costs.
The growing SAN-NAS convergence is part of a longer march toward more efficient use of disk-based storage resources and is similar to the trend of "virtualizing" server computers.
"We have not suffered any performance degradation," said David Reinker, an information systems manager at the agency. "In some cases, we've found it to be slightly faster."
BlueArc SiliconServer products allow for a SAN-like capacity of up to 256 terabytes in a single cabinet.
On Monday, the company unveiled a new product called "Titan SiliconServer," which also combines elements of NAS devices and SAN. The product can expand to greater capacities and adds features including the ability to increase system speed over time.
One sign of the push to merge NAS and SAN is the growing popularity of so-called NAS gateways. These are devices without disks of their own that can connect SANs with Ethernet networks. Worldwide revenue for NAS gateway products jumped 27 percent last year to $84 million, according to market research firm Gartner. Gartner expects annual NAS gateway revenue to grow 27 percent from 2002 to 2007, reaching $215 million in 2007.
By contrast, revenue for NAS equipment with integrated disk storage is slated to rise 16 percent annually during the same period, to $2.8 billion in 2007, according to Gartner. SAN gear revenue is expected to increase 11 percent annually in the period, to $10.3 billion in 2007.
The growing SAN-NAS convergence is part of a longer march toward more efficient use of disk-based storage resources and is similar to the trend of "virtualizing" server computers. A decade ago, much of the data used by servers was held on storage systems directly attached to those servers. But those isolated "islands" of storage translated into low usage rates--15 percent to 30 percent of storage resources, said Brian Babineau, an analyst at market research firm Enterprise Storage Group.
In recent years, networked storage has emerged to create larger pools of capacity that administers can carve up for improved efficiency. Use of SANs and NAS hit about 50 percent in the late 1990s and the early part of this decade, Babineau said. Since then, companies have turned to management software to increase the efficiency, and NAS gateways can help get rates up to about 70 percent, he said.
Ups and downs of SAN and NAS
A typical SAN set-up could include a number of disk arrays that are connected to a switch that, in turn, is connected to a number of server computers. SAN disk arrays are sold by companies such as Hewlett-Packard, IBM, EMC and Hitachi Data Systems, while SAN switches are produced by Cisco Systems, McData and Brocade Communications Systems.
SANs traditionally have been used with programs such as database applications that access data at the "block" level. In contrast, NAS gear provides data access at the file level, which is a higher-level way of organizing information. Block-level access is faster in a transactional database environment, because it requires fewer steps to locate bits, said Gartner analyst Roger Cox. Another key to SANs' speed is the use of the Fibre Channel interface, a technology that allows for data transfers at a rate of 2 gigabits per second (gbps). A standard has been approved that ups the data rate to 10gbps.
What's more, SAN capacity can be increased relatively easily by adding another disk array to the network. Drawbacks to SANs, though, include expensive equipment and difficulty in sharing files among users. Typically, sharing files in a SAN environment requires copying the data, which eats up storage capacity.
NAS boxes are dedicated computers that serve up files on a network. The devices are sold by companies including EMC, Network Appliance and Hewlett-Packard. Companies have used NAS gear to store files created in applications such as computer-aided design software or spreadsheet programs.
The file-level approach allows multiple users to share files without copying the data--for instance, several automobile engineers could work on the same new car design at the same time. NAS devices also are relatively simple to install, without the need for switches or network adapter devices. But NAS systems are generally slower than SANs. That's partly because NAS devices typically rely on Ethernet networks, which transfer data at a rate of 1gbps. What's more, NAS systems as a rule are not as expandable as SANs.
Bridging the data storage divide
NAS gateways, also known as NAS heads, are one way to combine the best of both NAS and SAN. These products offer greater ability to add capacity compared with a traditional, self-contained NAS storage system, said Jim Holley, EMC's director of NAS platforms. "If you start out with an appliance, eventually you're going to outgrow that appliance," he said.
Using NAS heads also allows companies to tap into unused capacity on SANs. And the products permit customers to trim the number of administrators needed to manage storage, said John McArthur, an analyst at market research firm IDC. McArthur said companies now have the option of replacing multiple NAS devices with a NAS head that links to a large-capacity SAN system. "If you want to simplify your environment, you want to consolidate," he said. "It increases 'vendor lock-in,' but it brings down operational costs."
In mentioning vendor lock-in, McArthur highlighted the limited compatibility of NAS gateway products. For example, Network Appliance said its NAS gateway products work with SAN equipment from Hitachi Data Systems and IBM, but not EMC's. IBM's NAS Gateway 300 works only with IBM's FastT storage servers and Enterprise Storage Server products. EMC's NAS gateway products are compatible only with EMC storage systems.
Aside from NAS gateways, another approach to bridging NAS and SAN technologies is to allow for both "block"- and "file"-level data access in the same device. This is a route Network Appliance has taken with its so-called unified storage products. Since 2002, the company has offered disk-based storage devices that allow for both block- and file-level data access.
More and more customers are using the products to handle data in both ways, said Chris Bennett, director of product marketing at NetApp. "It's common, and it's becoming even more common," he said.
A big, big file system
Still, another means to connect the SAN and NAS worlds is through technology that essentially creates a file system for SANs. For instance, IBM's SAN File System--launched last year--is designed to tie together servers in multiple locations over an Internet Protocol network and then allow the SAN to look and behave like a local file system, no matter where the data resides. Software keeps track of descriptive information, or "metadata," such as physical locations, file sizes and access permissions, that accompanies the actual content within the files, IBM said.
SAN File System is part of IBM's broader "on demand" computing initiative, which aims in part to provide information technology systems that smoothly handle spikes in usage. Sun Microsystems and Hewlett-Packard are working on related efforts.
IBM's SAN File System is more complicated to put in place than a typical NAS box. SAN File System software runs on metadata servers dedicated to it. In addition, software has to be installed on other servers using the system. But Big Blue said that the product can manage petabytes of data--one petabyte is equivalent to a million and a half CDs, IBM said.
"This is the ultimate of that (NAS-SAN) convergence," said Jeff Barnett, manager of marketing strategy for IBM's storage software unit. "A single NAS box is not going to scale to petabytes of data."
Perhaps not, but BlueArc said its new Titan product allows for up to 12 units to be clustered together for a total system capacity of 3 petabytes. Key to making its NAS products act like a SAN is the use of programmable silicon chips, according the company. BlueArc integrates a number of tasks usually reserved for software--such as protocols and network services--into so-called field-programmable gate arrays. That provides a speed boost that allows customers to run database programs or other applications that frequently use a block-level approach without a performance penalty, the company said.
"The distinction between SAN and NAS is blurred to some degree," said Geoff Barrall, chief technical officer at BlueArc.
The new Titan product has an architecture that will permit businesses to upgrade the speed of the product from 5gbps to an eventual 20gbps. What's more, the Titan supports larger file systems--up to 256 terabytes in a single file system--which can make for easier storage management, BlueArc said.
Two other smaller players in the storage world, Isilon Systems and Panasas, also have NAS products that are designed to let companies easily increase data capacity.
Although the NAS-SAN crossover has begun, the trend will increase with the emergence of iSCSI, Cox said. iSCSI is an interface technology that allows SANs to be built using less-expensive Ethernet networks. Another possible catalyst is the evolution of Ethernet bandwidth, which is expected to increase to 10gbps within the next couple of years, Cox said.
"Then you've got Ethernet on the same playing field as Fibre Channel," he said. "That's when all that gets real interesting."