Designing Storage Area Networks: A Practical Reference for Implementing Fibre Channel and IP SANs (2nd Edition)

Companies often find themselves in a reactive mode when attempting to meet the ever-growing demand for storage. When an individual file or application server exhausts its SCSI-attached storage capacity, an administrator may be forced to buy additional server/storage pairs simply to quickly provide more storage space. Over time, a data center may accumulate hundreds or thousands of servers, each with its own direct-attached storage. This spontaneous growth in the number of servers and the number of disks results in an expensive and often painful administrative headache. You need a much larger IT staff to manage large numbers of servers and storage devices, coordinate multiple tape backups, and maintain updates to servers and applications, and that requirement inhibits the alignment of data storage with corporate business priorities.

In addition, because each server is the exclusive owner of its attached storage, utilization of storage capacity throughout the enterprise may be out of balance. One server may have excess capacity while another exceeds capacity. Without a means to balance utilization, storage resources cannot be used effectively, resulting in unnecessary expense and administrative overhead.

The solution is to break the direct connection between servers and storage and to combine both on a common network. In this way, you can treat each element as a separate, manageable resource. You can assess server requirements independently of storage while sharing storage capacity among multiple servers. Storage area networking is thus the enabling infrastructure for consolidating both server and storage resources, reducing costs, and streamlining administration.

As shown in Figure 12-7, the spontaneous acquisition of servers and direct-attached storage imposes additional administration of multiple servers as well as multiple separate storage devices. Because storage is not shared, utilization may vary from server to server, resulting in inefficient use of your storage investment.

Placing both server and storage resources on a storage network enables you to reduce the number of servers and to consolidate dispersed storage capacity into a shared, highly available storage array. It also lets you manage more storage capacity with fewer administrators and use the familiar and more readily managed IP networking for high-performance storage access.

Depending on application requirements, the number of servers required to support database, Exchange, or other business applications can be reduced and consolidated onto more powerful platforms, as shown in Figure 12-8. Heterogeneous operating systems that previously required their own dedicated storage targets can now address their own volumes on a shared, highly available array.

Although storage consolidation on a SAN helps you to reduce the number of servers and individual disk arrays you must administer, it is not a panacea for all storage management issues. Without storage virtualization tools, you must manually configure the multiple LUNs on an enterprise-class storage target and assign each LUN to an individual server. Capacity management is now centralized, but capacity is not automatically allocated. In heterogeneous operating system environments, you must use LUN masking to conceal, for example, LUNs used by UNIX systems from Windows-based servers. For high availability, you must support dual data paths from the servers, redundant SAN switches, and alternative pathing on storage arrays. So even though the SAN is an enabling infrastructure for reducing server and storage overhead, it presents its own demands and issues. SAN-specific requirements demand due diligence in network design, but they more than offset the severe disadvantages of the spontaneous accumulation of direct-attached storage and servers.