The future of Internet SCSI (iSCSI) has been the subject of rampant speculation in the storage industry for the past 2 years. Although iSCSI represents the ultimate convergence of Storage Area Networks (SANs) with mainstream TCP/IP networking, some early-to-market products failed to establish sufficient end-user value to gain market acceptance. IBM's recent withdrawal from marketing its iSCSI-based TotalStorage IP Storage 200i array and Cisco Systems' lack of success with the Cisco SN 5420 switch have fostered an industry attitude that iSCSI still might not be ready for prime time. At the same time, however, the iSCSI-standards process is nearly complete, new wire-speed iSCSI host adapters and IP storage switches from other vendors are now available, and Microsoft announced support for iSCSI in its forthcoming .NET OS. These mixed iSCSI messages might well be confusing to the industry and to customers alike. However, the problem might not lie with iSCSI itself, but with how individual vendors have marketed the technology.
Some of the early vendor-marketing materials about iSCSI positioned the technology as a viable solution basically for low-end storage applications. Running an iSCSI device driver on laptop PCs, for example, would let laptops perform block-data backups over standard Ethernet networks. In part, vendors based this low-end market positioning on some products' limited performance. Vendors with wire-speed products, by contrast, have been positioning iSCSI as a solution for data-center and high-performance storage applications. The fact is, nothing inherently limiting in the iSCSI protocol restricts it to low-level or high-end applications. As with TCP/IP networking in general, customers get what they pay for, and those willing to pay more can get more in terms of functionality and performance.
The introduction of high-performance iSCSI host adapters (priced between $600 and $1000 each) challenges the vendor hype that iSCSI is much cheaper than Fibre Channel. Although iSCSI adapters are somewhat more economical than comparable Fibre Channel host bus adapters (HBAs), which range from $1200 to $2000, iSCSI adapters are still much more expensive than standard Gigabit Ethernet NICs, which range from $100 to $500. The real savings for customers, then, isn't in hardware acquisition costs, but in leveraging the IP infrastructure, expertise, and network-management tools already in place in their corporate networks.
Although the storage world doesn't expect enterprise-class iSCSI storage arrays to appear in the market until 2003, the availability of iSCSI host adapters and wire-speed IP storage switches will radically change the composition of SANs through the remainder of this year. Why? Because now customers will have the flexibility to deploy servers anywhere in the IP network instead of tethering servers to dedicated Fibre Channel switches. By using more of their familiar IP infrastructure and support staff to implement SAN solutions, customers can expand their storage networking applications without simultaneously hiring more Fibre Channel fabric experts. When iSCSI storage systems come to market, customers can simply integrate them into their existing IP SAN, side by side with Fibre Channel arrays and hosts that storage switches have already brought into the IP SAN. As Carlson Companies and other customers have shown, implementing IP-based storage networks is viable today, even as the market continues to introduce new IP storage products continue. (For stories about the Carlson Companies in Computerworld and Byte and Switch, respectively, see the URLs below.)
This heterogeneous iSCSI migration lets customers select the best-in-class Fibre Channel storage and iSCSI adapter products without being locked into a specific technology solution. Customers can set their own pace for integrating IP and Fibre Channel products while taking full advantage of the functionality and performance that both offer.