Storage networking, IP style

The recently ratified iSCSI protocol provides another way to build and link storage-area networks.

The storage networking specification, which uses standard IP networks to carry data, has been heavy on concept and light on usability until recent months. In February, iSCSI earned the Internet Engineering Task Force's seal of approval. Then Microsoft Corp. lined up behind the protocol, releasing an iSCSI driver for Windows PCs and servers. And storage vendors such as EMC Corp. have incorporated Gigabit Ethernet cards into storage arrays to communicate directly via IP. Thus far, iSCSI implementations have been modest and departmental in nature. But Ahmad Zamer, chairman of the Storage Network Industry Association's iSCSI group, said he believes the killer application for iSCSI will be remote data replication and storage. In that respect, iSCSI is best suited for asynchronous replication, he added, noting that synchronous mirroring could cause delays over a wide-area network. John Joseph, vice president of marketing at EqualLogic Inc., a maker of iSCSI storage arrays, noted that most of the company's installations thus far have been within a single site. But he said future functionality would enable the company to take on disaster recovery and asynchronous-mirroring scenarios. He said EqualLogic plans to provide multisite capability for linking storage facilities over distances approaching 200 to 300 miles. Some industry executives, however, contend that iSCSI is not well-suited to high-performance environments — relational databases, for example — and raises security concerns common to the IP world. Zamer, however, pointed out that the iSCSI specification requires the IPSec security protocol.