NAS and SAN: Feeding the Hunger for Storage
With a seemingly insatiable demand, organizations are scrambling to increase storage capacity. But when faced with the prospect of adding hundreds of gigabytes or terabytes of storage and hiring more administrators to manage the storage each year, organizations find it impractical to keep adding conventional server-attached storage. Instead they are turning to new storage approaches: network attached storage (NAS) and storage area networks (SANs).
While often portrayed as competing technologies, NAS and SANs are actually complementary, argues Bob Katzive, vice president at Disk/Trend (www.disktrend.com), a storage trends research firm. Of the two storage technologies, NAS promises to deliver more benefits in the short term; while the more immature SAN technology will take a few years to deliver its promised benefits.
Driving the need for storage are the insatiable demands of workers and organizations for information. Data warehousing, data marts, OLAP and a new generation of analytical applications consume enormous amounts of storage. New applications that take advantage of imaging, graphics and multimedia/streaming media have a voracious appetite for storage.
The largest storage consumer could be the Web, with its e-commerce and information publishing applications. Web sites, Web caching, Web site replication and the need to capture and track the activities of Web site visitors requires massive storage capacity. Organizations are laying plans to capture and mine enormous volumes of data for purposes of personalizing the Web experience, cross selling and customer relationship management (CRM).
Further complicating the storage problem is the fact that different applications and situations have different storage requirements. In addition, there are storage administration and management issues to be considered.
"There are a lot of questions about storage. Some applications need storage tightly coupled to the server. Other environments can use separate storage. Then there is the question of how to centrally manage all the storage. You want to dynamically grow the storage without adding more servers," explains Richard Guerra, director of engineering at Compaq Computer Corp.’s enterprise appliance servers division.
Through the conventional approach, organizations simply added storage to an application or file server. They would fill up storage bays within the server and add adjacent storage cabinets if necessary. The servers were attached to the network, and the server and its applications handled all access to the stored information. With the rapid growth in storage requirements, however, this approach quickly becomes unmanageable. Organizations can experience an uncontrolled proliferation of servers with attached storage, greatly increasing the workload on system administrators and resulting in the need to hire more administrators.
NAS and SAN solutions offer storage approaches that reduce or eliminate the problems encountered when organizations try to scale up server-attached storage. Both NAS and SAN separate the storage from the server. This allows for large amounts of storage capacity to be added and shared without the need for additional servers and a corresponding increase in system administrators.
NAS and SAN are very different approaches, however, each with its own protocols and advantages, and each at a different stage in technological maturity. In general, "NAS is storage for client systems while SAN is storage for application servers," Compaq’s Guerra points out. NAS focuses on file sharing, with NFS (Network File System) and CIFS (Common Internet File System) as key protocols. NFS is the file-sharing standard in the Unix world. CIFS, formerly known as Server Message Block (SMB), is the Windows equivalent.
With its NFS/CIFS foundation, NAS is optimized to provide file-level access to information, which is why it is considered advantageous for client systems. A NAS device, often referred to as a storage appliance, can be used to consolidate numerous file servers, managing access to volumes of file-level information, explains Howard Alt, CTO of network storage at Sun Microsystems Inc. SANs, by comparison, use block-oriented protocols that are favored by databases and database applications. Block-oriented protocols enable more granular access to data.
While SANs have received more attention in the industry trade press and at conferences, NAS has quietly racked up some impressive numbers. International Data Corp. (IDC, www.idc.com), projects the NAS market will grow 54 percent per year through 2003, when it will surpass $5 billion in sales, up from $540 million today. Fuelling the demand for NAS is "the size of the desktop data explosion, the value of client data and the demand for data security in the client LAN space," observes Robert Gray, IDC research director of worldwide storage systems.
The current crop of NAS storage devices is aimed at the client market. For example, The Hill School, a prestigious private school in Pottstown, Pa., installed three NAS devices from Hewlett-Packard Co. to serve up CD-ROM- and DVD-based information directly to its users. "Previously, we were copying files to magnetic hard drives on servers to make it accessible to students and teachers," explains Rick Bauer, CIO at The Hill School. Demand for more information, however, put the school in an awkward position. "We were really chewing up server storage space," he recalls.
The school introduced one HP NAS CD-ROM device as a test. "It was breeze to configure. Just plug it in, turn it on, and give it an IP address and NETBIOS name. You plug it into the network just like you would a printer," Bauer explains. It was so easy the school added two more. If the school opted for a non-NAS CD-ROM storage device, "then we would have had to attach it to a server, which meant loading a couple of drivers and setting up volumes," Bauer explains. The NAS device was not only easier to set up, but it doesn't eat up any server processing resources.
Bauer's experience with the HP NAS appliances has him thinking about NAS for hard disk storage, as well. All he is waiting for is the right device, one offering hot-swap drives that he can snap in or out without bringing the down the device.
NAS is often suggested as an alternative for small organizations and departments; while SANs are generally regarded as an enterprisewide solution. NAS devices start as small as 50 GB, Compaq’s Guerra says, although there is no reason they cannot scale up to a terabyte or more. The smallest SAN implementations, however, start with hundreds of gigabytes and scale to multiple terabytes.
Also, NAS is considered a simpler technology to implement. NAS devices are as sophisticated as plug-and-play storage devices can be. The key protocols -- IP and NFS/CIFS -- are already widespread and stable, compared with SAN technology, which is still evolving. SAN implementations require considerable support from vendors and system integrators. "If an organization does not have the IT expertise for SAN, they can use NAS and the existing network in the interim," Guerra suggests.
The distinction between small and large organizations may not be the most accurate way to view NAS and SAN. It comes down to the type of information being stored and the nature of the access. Large organizations can use NAS storage to consolidate file servers as well as smaller companies. Those same NAS devices could also connect to the enterprise SAN out the back end.
"In the future, you can have NAS as a gateway pulling down data from the SAN," IDC’s Grey notes. In that situation, users would access data files across the LAN from a NAS device. The NAS device would connect through its back end to the SAN, from which it would pull down the data it will make available to client systems. Most databases, Grey adds, also work under a file system manager, which would reduce the need to support block-oriented data.
For now, most organizations will turn to NAS devices for server storage consolidation or for supplemental storage. With its native IP and file system protocols, NAS devices also are obvious choices for bolstering storage capacity for a Web site.
"NAS is good for storing traditional files, so most organizations will use it to consolidate user directories," Sun’s Alt suggests. In one scenario, an organization would implement a NAS storage device for a set of servers on the LAN. This will eliminate the need to manage storage at each of the servers. With the cost-per-megabyte of storage falling each year, "the real cost of storage is in storage administration," Alt continues. In this scenario only a single NAS device must be administered.
There are, however, some issues to resolve with NAS. For instance, backup presents a challenge. The NAS device must be backed up locally, otherwise the organization risks swamping the network during a large backup. Where the NAS data is provided from the SAN, backup is handled by the SAN and isn't a NAS issue.
Organizations may also have to face some NAS configuration decisions, such as whether to build the storage capacity out of fewer large-capacity disks or multiple smaller-capacity disks, Alt says. Larger disk drives deliver the capacity efficiently, but many smaller disk drives allow organizations to achieve more input/outputs per second due to multiple drive spindles. The latter is probably preferable for handling the high volume of hits on a Web site.
One thing that is not a concern is the operating system. "The nice thing about NAS is that it hides storage behind an open protocol. I use NAS regularly with Unix and NT systems, and it works as well with NT as with Unix," Alt reports. Disk/Trend's Katzive adds, "Because NAS uses a limited set of network protocols, it encounters fewer open systems problems. It will work equally well with Unix, NT or Novell."
As organizations wrestle with unrelenting demands for additional storage capacity at both the client and server levels -- projected to increase from 100 percent per year storage growth today to 200 percent per year in only a couple of years, according to Forrester Research (www.forrester.com) -- they will need every tool at their disposal. In addition, as IT organizations stretch resources, they have to take advantage of every strategy that can reduce the need to hire additional system and storage administrators. As a result, the overall enterprise storage architecture of the future will likely include a mix of NAS devices connected to a SAN as organizations attempt to consolidate and centralize database, application and file storage at all levels.