Network attached storage (NAS): NAS is a storage device attached to the network like any node. NAS servers can access data directly from only the devices they control.
Storage area network (SAN): A SAN is a discrete network of data-storage devices and servers that allows unrestricted exchanges of information. SAN servers can access data directly from any device and do not transfer data over the LAN.
If you are planning to move your business to the Internet, you have probably carefully considered which Web server to purchase, balanced the benefits of various Internet-commerce packages, and determined how to market your products to the online community. But have you thought about where to put the gigabytes of data with which your Web store will suddenly bombard your network?
Since 1996, the best option for network storage has been network attached storage (NAS). A NAS is a platform-dependent server or pool of servers devoted to storage that adds gigabytes of disk space to your network as easily as adding a new printer.
But because each NAS box must be managed separately, it may become too complicated an approach for storing the data of a growing enterprise; it is very well-suited for workgroup or department storage, handling about five servers.
Late in 1998, a new approach to network storage, called a storage area network (SAN), entered the market, promising to eliminate the limitations of NAS.
A SAN is a network separate from the LAN with unlimited storage capacity. It is an ideal centralised storage system for an enterprise. A SAN is based on a very simple principle: any server can exchange data directly with any storage device, without the limitations imposed by NAS architecture. A SAN not only will hold the contents of your Web servers, extranet, and intranet, but it also will let you manage all of that data from a single, central point.
The most important difference between a SAN and NAS is how the storage devices and servers exchange data. A NAS system is based on SCSI point-to-point connections between a server and data devices. Each NAS server connects to the LAN with standard network protocols, such as TCP/IP, and monitors the data transfers between data devices and other nodes on the network.
However, this limits access to the data through the NAS dedicated server only; when other nodes need to access the server, the data must travel through the LAN, increasing the LAN's traffic as the data flows from server to server through the LAN.
A SAN obviates the problem of conflicting bandwidth demands with its any-to-any connections. Unlike a NAS system, a SAN does not use an existing LAN to transfer data between servers and devices; data travels over a SAN without affecting the LAN.
On a SAN, a full network backup can run without harming applications traffic, which is a tremendous advantage to companies running a round-the-clock business. Also, because a SAN uses an extremely fast connection protocol, it dramatically reduces the time it takes to complete a backup.
In addition, a NAS system's SCSI connection limits the number of devices that can be attached to a server to 16 for each SCSI controller. To increase storage -- and management headaches -- beyond that requires adding a new connection to the server or a new server. And SCSI is slow -- in a NAS system, the maximum data transfer rate is generally 80Mbits/sec (bps). But the new Fibre Channel standard for SANs, Fibre Channel Arbitrated Loop (FC-AL), can connect as many as 126 devices and transfers data at 200Mbps, easily outperforming a NAS system. FC-AL is 10 times faster than Fast Ethernet, and future versions promise to increase performance to speeds four times faster than the current mark.
Another constraint of NAS systems that is addressed by a SAN is the possible distance between a server and storage. Companies often prefer to isolate their data in a room with restricted access, but the 25-metre limit that SCSI imposes gets in the way. A SAN equipped with FC-AL expands the possible distance from 25 metres to a whopping 10 kilometres for a single loop, which allows for a respectably safe space between servers and data devices.
For companies with increasing storage needs, moving to a NAS system makes good sense -- it is simple to install and affordable, and it is based on mature technologies. But because NAS systems are costly and complicated to manage, it is best-suited for departmental storage. If you need to store terabytes, you should move your data to a SAN, which can be managed and expanded independently from the servers on your LAN.
If you already have installed a NAS system, you can upgrade your existing equipment to a SAN. A SAN uses data routing to manage the connections to different devices. Data routing allows a newly implemented SAN to connect to existing networks and NAS systems; a SAN can convert data traffic from one protocol to another with a router, connecting a Fibre Channel device to a SCSI device.
For example, using a Fibre-to-SCSI bridge, you can inject new life into an existing SCSI storage device and add its capacity to the pool managed by a SAN. Routers are an important component of SANs because they foster reusability of current media devices and interoperability with existing networks. They also let SANs interoperate with the major protocols in an enterprise, including ATM and Ethernet.
Also, having a separate network for storage removes the load of data-only applications, such as a backup, from your LAN. More importantly, a SAN has to manage only heavy data traffic and not interactions with users, so it can be optimised to provide the bandwidth capacity and the flexibility such heavy traffic requires.
Last year, big-name storage companies -- including StorageTek, Veritas, and Legato -- announced plans to build SAN-compliant devices and management software. By December 1998, several major networking players -- including Compaq Computer, 3Com, and Sun Microsystems -- announced their intentions to develop products such as hubs, switches and servers with which to build a SAN. Some of the vendors already have released SAN products, and others plan to make theirs available early this year. However, implementations will probably vary greatly between vendors.
For example, some storage vendors, including StorageTek, support a SAN infrastructure that can be open to various present and future technologies. StorageTek is promoting Virtual Intelligent Storage Architecture, or VISTA, an open, information-centric approach to SAN that defines separate layers for each component.
However, "open" has different meanings depending on the SAN vendor. StorageTek's approach is flanked by reputable names such as Compaq, Sun, and Veritas, that agree storage should connect seamlessly to a clearly defined, common framework.
But other storage companies, such as EMC, favour a storage device-centric approach, in which compatibility between servers and existing storage management software should be developed. To further confuse matters, even though most SAN vendors support the FC-AL standard, IBM has proposed a different SAN solution based on Serial Storage Architecture, or SSA.
Do you need a SAN? Analysing your company's bottom line will answer that question. If your company pays dearly for downtime, waiting for backups to complete, a SAN can reduce that cost dramatically. A SAN can also decrease the costs of managing your storage devices. If you keep increasing the bandwidth on your LAN without apparent relief, a separate network for data could be the break that your company needs. A SAN also offers a reliable method for handling your disaster/recovery issues. Even if your company does not currently show any of these symptoms but the amount of your storage is ballooning, you may want to consider a SAN to prepare for the future of your network.
Technology analyst Mario Apicella has held several IT positions including 11 years as a system programmer and database administrator. He has worked with IBM mainframes, Cobol, NetWare and Windows NT. At InfoWorld he focuses on databases and application development. Associate editor Leslie O'Neill has been a high-tech journalist for several years. She currently specialises in enterprise networking-related articles for the InfoWorld Test Center. Send comments or questions about this article to firstname.lastname@example.orgAT A GLANCENetwork attached storage (NAS)NAS is designed to provide hassle-free, reliable storage expansion for small or departmental networks. A NAS unit attaches to an existing network and exchanges data with the network's client applications through the LAN. Several self-contained products that include a server optimised for data transfer and storage media are available, and some of them offer the option of adding a tape-based backup system to the unit.
Pros: Easy to install; affordable; based on mature technologies, including SCSI.
Cons: Limited performance and storage capacity; increased management costs due to proliferation of devices; limited distance between controllers and storage devices; server-controlled access to the stored information.
Storage area network (SAN)
A SAN is a new approach to network storage that promises to separate storage management from server management with a discrete network devoted to storage. A SAN also supports diverse network protocols, operating systems, and applications. Most SAN vendors, including StorageTek and Compaq, are converging on a Fibre Channel-based SAN. However, IBM, a notable exception, has proposed a proprietary architecture.
Pros: Open architecture compatible with existing infrastructure; removes data-transfer overhead from the LAN; permits unrestricted data transfers among servers and storage devices; facilitates backup and disaster recovery.
Cons: Complicated to install; based on immature technologies; expensive.
Powerful storage companies pioneer SAN
Several big names in the storage and networking industries, as well as a few small companies, are leading the SAN evolution. Hardware and software vendors are beginning to release products that support SANs, including servers, drives, switches, hubs and management software.