Enterprise storage consolidation is on the rise because it increases efficiency, decreases redundancy, and simplifies management. You've probably heard how important enterprise storage is to businesses, and you might have also heard of two hot storage technologies: Network Attached Storage (NAS) and Storage Area Network (SAN). NAS devices are storage appliances—big, single-purpose servers that you plug into your network. These appliances perform one task, and they perform it well: They serve files very fast. The capacity of large NAS appliances is in the terabyte range. SANs are multiserver, multistorage networks and can grow larger than 400TB. A SAN acts as a secondary network to a LAN. Every server that needs access to the SAN has a fibre channel connection to the SAN. This secondary network relieves the main network of massive data transfer loads because backup traffic occurs between storage devices inside the SAN.
NAS appliances and SANs are substantively different in their capabilities. A NAS appliance can perform certain functions that a SAN can't perform, and a SAN offers scalability. Let's explore the underlying technologies of NAS and SAN and look at scenarios in which you would want to use one technology instead of the other.
The Motive for NAS and SAN
The traditional approach to storage, which the model in Figure 1 shows, creates several storage difficulties. Three or more years ago, the only storage approach available was one in which you directly attached storage (e.g., hard disks, RAID arrays, tape media, optical disks) to each server. The server processed all client requests for files.
This storage approach results in several administrative problems. First, the traditional storage model is resource-inefficient because the storage space exists in isolated pools. One server might be out of storage space, yet another server might have 100GB of free disk space. Traditional server storage also creates unplanned, unnecessary redundancy—duplicate copies of the same file might be on several servers. The duplication is inefficient and makes information management and collaboration difficult. Also, when servers on the LAN supply the storage, backup traffic ties up network resources.
A final problem is that servers can't share data across platforms because one file system (e.g., NTFS) can't read another file system type (e.g., UNIX). Some third-party applications can perform file-system translation, but these applications are difficult to manage. Therefore, the traditional server storage model tends to waste valuable resources and causes administrators to take extra time balancing the storage load, managing redundancy, and avoiding bottlenecks. NAS and SAN are the solution to many administrators' storage headaches.
What's the Difference?
The difference between how a NAS appliance and a SAN function is subtle. A NAS appliance uses an IP protocol to serve files to clients, and a SAN uses a SCSI protocol to serve data blocks to servers. A NAS appliance resembles a network server from which clients can obtain files, but a SAN appears as extra storage for each server. NAS appliances send files on request; SANs grant direct access to disks.
Another way to look at the difference is that Ethernet connects a NAS appliance to the entire network but SCSI or fibre channel connects servers to a SAN. Therefore, NAS appliances offer high performance and the capability of serving files to heterogeneous servers.
SANs improve on the traditional server-based storage model and provide high-speed data access. A SAN joins several storage devices into a storage pool that has partitions, and each server on the main network has an assigned storage partition. Therefore, SANs use the traditional storage topology model in which each bit of storage sits behind a server. However, SANs let you quickly retrieve data, easily repartition and assign storage space, and relieve the LAN of backup traffic.
A NAS appliance is a dedicated storage server (i.e., storage array) that you plug into your network, as Figure 2 shows. Clients send file requests directly to the NAS appliance, bypassing the general-purpose servers on the network. Most NAS appliances contain an embedded, stripped down, and optimized OS, such as a UNIX kernel, that only serves files. The appliance also contains an I/O processor and a file processor.
A NAS appliance has a proprietary file system and can serve files across platforms because it can read all major file systems. The dedicated microkernel dramatically decreases complexity and increases reliability. NAS appliances also have the ability to share a data instance on multiple application servers, which makes NAS a stunning cross-platform, collaborative tool. NAS appliances are cheaper than standard servers, so you can buy a NAS appliance for Windows NT at a very low price. Therefore, a NAS appliance's cost for performance makes it an attractive package.
NAS appliances can also take snapshots, pictures of the client or server directories that store the data. Taking snapshots is more common in the mainframe world than in the NT world, and is different from saving data. Snapshots are records of where the data resides, not copies of the data. Saving a duplicate file copy can take several megabytes, but a snapshot takes only a few bytes. Snapshots can restore an earlier file version automatically. For example, you can configure a NAS appliance to take a file snapshot every 30 minutes. Then, you can use the snapshot to recover a file that you've edited and saved but which you need to retrieve the earlier file version of.
SAN in Brief
A SAN is different from a NAS appliance because a SAN is a network, which Figure 3 shows. Network architects design SANs to meet the needs of specific businesses. (For more information about SAN design, see Tom Henderson, "SAN Topology," page 91.) Let's look at the general characteristics of the SAN.
SANs are secondary storage networks that include standard components such as servers, multiplexers (MUXs), bridges, and storage devices (e.g., tapes, disk arrays). Each server on the main network uses a SCSI or fibre channel connection to the SAN, so every server receives high-speed access to the data storage. Each network server treats its allocated SAN space like a directly connected disk, and the SAN uses the same communication protocol that most servers use to communicate with their attached disks. The SAN model pools the storage into one network, and each server on the LAN has a connection to the entire SAN. The storage sits behind each server, so unlike the NAS model, clients must send file requests to the servers to receive SAN data.
SANs provide the advantage of storage allocation flexibility. SANs have repartitioning and management tools, so when administrators want to reallocate storage space from one server to another, they simply repartition the SAN. Repartitioning is easy because you assign storage space to a network server instead of directly connecting storage space to a network server. Another SAN advantage is that it relieves the LAN of backup traffic. Because the storage resides on a secondary network, the backup traffic doesn't need to travel over the main network. This advantage is a major selling point in many businesses.
The Final Conflict
NAS appliances offer performance and reliability at a low cost. NAS appliances are excellent devices for collaboration and data storage, especially in heterogeneous computing environments. Yet, NAS appliances can send only files, not data blocks, which limits their ability.
A hybrid solution for data storage is to set up a SAN, then connect it to a NAS appliance. When you make the NAS appliance the gateway to the SAN, you receive cross-platform access to all the SAN's data. Technically, you create a big NAS appliance that uses the SAN for additional storage space. Unfortunately, this solution yields NAS's limitations, such as the lack of direct disk access.
SANs provide disk space to servers, and how the servers use that disk space depends on the servers' file systems. Although SANs can't provide cross-platform support, they provide the direct disk access that many major applications, such as Microsoft Exchange Server, require. SANs are very scalable, and they offer enormous flexibility. For example, you can add 10TB of disk space to a SAN and designate the disk space to the specific server that needs it.
Don't let SANs' promise to free your network of bottlenecks fool you. This traffic relief comes at a high price because a SAN adds many additional and costly resources (e.g., servers, disk arrays). A SAN also requires a connection to each network server.
NAS appliances and SANs supply enterprise storage and offer storage consolidation. However, they provide different services, advantages, and limitations. The solution you choose will depend on your needs.