CertiGuide to Network+  9  Chapter 0110: Network Operating Systems (NOS)

Pop Quiz 0110.10 1 XII  Comparing SAN and NAS

Sometimes two or more servers are tied together to make them look like one larger server in very high performance networks, called System Area Networks or Clusters. This allows the servers to communicate extremely high volumes of data between them without the usual overhead of TCP/IP networking present in NAS or multiple file server based solutions. The end results of this are that more CPU resources are available for use by applications.

Additionally, depending on the complexity of the SAN technology provided by the vendor, some SAN’s allow for sharing of resources such as RAM and even CPU microprocessors between machines. This allows entire servers to share processing jobs, with work swapped among processors to even out the workload among them. The process of evening the workload among servers is referred to as load balancing. The advantage of load balancing is that application performance is more predictable – you’re less likely to encounter a situation in which users assigned to “the fast server” get fast response time and users assigned to the more heavily loaded “slow server” have to wait substantially longer for their applications to run. A SAN would move some of the file server activities from the “slow server” to the “fast server” in an attempt to provide more uniform response time.

Another benefit of this approach is scalability. When you deploy a major application to more and more users, your server performance may begin to suffer. In the traditional approach, you would replace the existing server with a larger, faster one. Using a system area network, you would merely add another server to your existing server, preserving your initial hardware investment. This ability to add rather than replace can be very handy when your application performance demands tax the capabilities of your vendor’s highest performance, highest capacity machine. With the traditional approach, you’d have to look for a bigger machine from another vendor, and be out of luck if a faster machine was not available.

Yet another benefit to a SAN, which is also a benefit of NAS, is increased server fault tolerance. We’ve already seen the increased fault tolerance provided by using RAID technology for disk storage.

Since multiple servers can access the same disk storage (RAID or otherwise), it is possible to keep a backup server on-line, which can quickly take over the work of any server that fails, much like disk mirroring allows for applications to continue, using the data on the mirrored drive, if an error occurs on one of the drives.

A downside to using a SAN is that applications usually need to be written specifically to support SAN technologies, and not all applications are. You are more likely to find support for SAN’s in high-end “enterprise class” applications such as databases like Oracle.

To connect servers together into a SAN, you use special purpose high-performance adapters called system area network cards, rather than standard Ethernet cards. You also need special operating system and/or application software (depending on the specific SAN platform you are using), which understands how to communicate among the devices on the SAN.

Also, Microsoft is promoting a new API called “Winsock direct” which aims to standardize how Winsock-compliant applications interact with Winsock-compliant devices on SAN’s.

A low-cost Linux-based hardware-independent clustering solution, called Beowulf, is becoming increasingly popular in scientific communities, allowing essentially commodity hardware such as Pentium Pro based PC’s to be used as servers in a clustered SAN environment, allowing the processing power of multiple PC’s to be grouped together for parallel processing of scientific data.

Pop Quiz 0110.10 1 XII  Comparing SAN and NAS