SOME OF THE MAIN TOPICS IN THIS CHAPTER ARE
All the components that make up a computer, from the CPU to memory, the bus, and others, are rapidly expanding in speed and capacity and probably will continue to do so for many years to come. One area of computer technology that while continuing to evolve is doing so at a somewhat slower pace is storage devices attached directly, such as disk and tape drives. For desktop workstations that use limited storage, this is not really a problem. For servers that offer file services to other computers, this is a serious issue with which administrators must contend. As the need for storage continues to grow, it is pushing the limits of the most typical server storage interface today: the SCSI ( Small Computer Systems Interface) interface. And because most large servers use RAID arrays, even more disks are needed to store data due to the mechanics of disk mirroring, striping, and so on. The SCSI parallel architecture, using multiple wires and intricate clocking, can work only over short distances. The different levels of SCSI that exist today have different distance limitations as well as limitations as to the number of devices that can be attached to a SCSI cable. Because of these physical limitations, it is apparent that eventually you will run out of PCI slots to hold SCSI cards, and that all the storage you can fit within the limits of the current SCSI capabilities will be exhausted, as shown in Table 11.1.
One practical solution would be to get another server and divide the chores. But for some very large servers (or clusters) that provide Internet services, or other applications that require a large amount of storage, the physical SCSI architecture is now seeing its last days. You can only connect so much storage to a server, given the distance limitations of SCSI technologies and higher disk capacities. Eventually, disk storage space that you can connect to a server will become finite, and a bottleneck. Thus, parallel SCSI does have limitations. The new Serial Attached SCSI (SAS) standard, which combines compatibility with Serial ATA (SATA) hard disks with the management features and performance of SCSI, is expected to eventually replace parallel SCSI implementations at the server level.
Newer technologies are needed in just a few years to satisfy the need for growing, reliable storage. One is the emerging InfiniBand technology, and the other is the proliferation of Storage Area Networks (SANs). As enterprise data begins to be measured in terabytes now instead of gigabytes, there simply must be a better way to make data available to one or more large network servers. Currently, that methodespecially for large networks in which downtime is very rarely toleratedis the Storage Area Network (SAN). Another method of expanding data storage by using networking technologyoften used in small to mid-sized network environmentsis the Network Attached Storage (NAS) device (sometimes called a network appliance), which you will also learn about in this chapter. NAS devices are attached to a LAN along with client and server computers. SANs are usually connected to the larger servers using a separate network. There are exceptions to this rule, of course. Some enterprise networks still use NAS because it was "there first." There is no need to replace what works with a more expensive technology just because it exists. In this chapter you will find compelling reasons for using NAS, SANs, or possibly both in your network, depending on your requirements. Both of these technologies have specific features that make them useful in different environments. You will find that some networks can make use of both technologies, again, depending on the data requirements for the network. First, let's examine the most widely used technology for attaching storage devices to a server: the Small Computer Systems Interface (SCSI). We'll also cover the limitations of this legacy technology in large enterprise networks. |