Introduction to SANs: Technology, Benefits, and Applications - Page 6

There is much confusion surrounding the differences between a SAN and a network-attached storage, or NAS device. A NAS device is a self-contained, autonomous storage appliance connected directly to a network. Unlike SAN, which operates at a block level, NAS is a file oriented storage solution, which functions like traditional file servers in an IT infrastructure.

What is NAS?
The storage resources of a NAS device are accessed via mapped network drives which talk to servers and clients through the NFS (network file system) or CIFS (common internet file system) protocols, which are run over the IP protocol. The following diagram illustrates how NAS serves clients.

NAS Diagram

NAS is very simple to set up and maintain. The thin server technology used by most NAS devices is easy for network administrators of any skill level to understand and is very reliable. However, NAS is not a replacement for SANs.

For one thing, as more NAS devices are added to a network, their sheer number can become unwieldy. While some NAS devices have the ability to scale to several gigabytes in size, a well-designed Fibre Channel SAN can scale to accommodate terabytes worth of storage. As NAS use proliferates throughout a network, an IT manager has no option but to add more NAS units. Although an individual NAS device is easy to manage, it is extremely difficult to globally set parameters like security levels or application access to all NAS devices simultaneously throughout a corporate network, a crucial flaw that we wish NAS vendors would fix.

Overall, NAS is a great solution to centralize files on a workgroup level, but for companies that want enterprise-wide storage resource centralization for their distributed systems, a SAN is the only viable answer.

Differences Between SAN and NAS: File Level vs Block Level Protocols
The basic technical difference between SAN and NAS lies at the communication protocol level. While NAS communicates via NFS and CIFS Fibre Channel, SANs transmit data over FCP (Fibre Channel Protocol).

As the name implies, File Level protocols transfers data from storage units to servers in the form of files. Unlike a SAN, NAS solutions have their own file systems, which they manage independently within their storage enclosures. Although thin server units in charge of their own file system management and user authentication are convenient and typically easy to manage, a NAS unit can become a serious bottleneck in high transaction environments. (See our discussion of NAS performance below).

In the case of Fibre Channel, data is delivered in device blocks the same way an external RAID array would deliver data to an attached server. In a SAN environment the middleman, a thin server unit that typically manages NAS, is eliminated. Furthermore, since a SAN doesnt have to use IP to transfer data to the servers, the conversion of data blocks into IP packets (which is done in NAS) doesnt have to take place, and this reduces latency.

SAN and NAS Together
So, what is the role of network attached storage in conjunction with a SAN? Contrary to many beliefs, NAS is not replaced by SANs. Part of the confusion between these two technologies is due to the fact that both SANs and NAS have the ability to centralize data. NAS allows users to share data as a single files in a single data repository, leveraging networking technology (Ethernet) that virtually every company has implemented over the past decade. Since NAS moves storage traffic off of the SAN and onto IP networks, IT managers can use NAS to reduce the pressure in scaling their SAN infrastructures beyond their existing means, saving some infrastructure cost, though SANs by nature are extremely scalable. In its current state, SANs are best suited to serving the needs of transaction-intensive application servers, while NAS (since they dont require extra infrastructure) are best suited for the storage needs of low end workstations and desktops.

If NAS is cheaper to implement and has similar storage centralization capabilities, why should we bother to go through the painful process of SAN implementation?

Answer: Performance.

For performance reasons, most IT managers would not want to run disk-intensive applications on a NAS device across the network. Some applications, such as databases, work best when the database application has direct high-speed access to their directories, data and log files. NAS works well for file server access in most situations, but an organization should still use traditional server-attached or SAN storage subsystems to run disk-intensive applications such as transaction processing.