Why IP SANs?

Ultimately, the question of why IP storage seems to be answered by the burgeoning requirements for block-level storage scaling. When introduced to market in the late 1990s, SANs reflected a continuing bifurcation of the storage market. SANs provided a scaling mechanism for block-level applications such as databases, while NAS provided " elbow room" for files. Recent events, such as the announced substitution of an object-oriented database for a file system in Microsoft's forthcoming operating system ”code-named "Longhorn" ”combined with similar structural changes in popular databases from IBM (DB2 Open Edition) and Oracle Corporation, suggest that scaleable block architecture may become de rigueur for all storage in the future.

This is not to say that IP SANs provide a "one size fits all" solution to application storage requirements. The very proliferation of protocols and topologies suggests a diversified solution set that may map to the different requirements of different applications. Just as different RAID levels are preferred by different elements of an Oracle database to obtain optimal performance, so, too, different storage protocols and architectures may be required by different applications. For example, for streaming data applications, such as tape backup, iSCSI may actually be a better match than Fibre Channel, owing to the support in Ethernet for "Jumbo Frames ." Independent tests have verified that the use of Jumbo Frames (extended Ethernet frames that range in size from the standard 1,518 bytes up to 9,000 bytes) can deliver a 50 percent increase in throughput with a simultaneous 50 percent decrease in CPU utilization in streaming tape applications. Alternatively, and this is likely to raise arguments, applications requiring a specific class of service support (an available but rarely implemented capability of FCP) might be well- served by a Fibre Channel interconnect.

Ultimately, the decision over which protocol to use with which application must consider a number of variables including application performance requirements, cost, available support resources, management and security requirements, and data growth rates. Without minimizing the differences between iSCSI and FCP, it needs to be clearly understood that interconnecting servers and storage using either of these protocols is largely irrelevant to the SAN value proposition. Plumbing is just plumbing. Intelligence ”some would say, management ”is required to create an infrastructure from the aggregation of disk drives , cables, switches, and HBAs. Without it, you do not have an infrastructure, just a morass of hardware and cabling.

What SAN plumbing does provide, in addition to protocol-specific features, is the foundation for building resilience and provisioning efficiency into storage. The IP storage model arguably provides a better foundation because of its tenure. We have a better understanding of techniques and methods for designing IP LANs to achieve optimal performance and to minimize risk, knowledge that can be transferred readily to build high-performance, fault-tolerant IP storage networks. There is no equivalent tenure in Fibre Channel.

Added to this is the ubiquity of IP. Storage, like any other application, gains part of its value from its accessibility by those who need it. IP networking, in its many manifestations , enables broad geographic accessibility.

Ubiquity and tenure also conspire to make IP storage networking components cheaper, product development times shorter, and vendors coffers bigger. At the time of this writing, the research and development budget at Cisco Systems was greater than the budgets of the top five Fibre Channel switch- makers combined. This fact bodes well for the ability of IP SAN technology to keep pace with changing storage technologies and changing business requirements. It is also a practical explanation for why IP has buried virtually any rival technology that has crossed its path .