To hear the spec engineers tell it, the creation of the iSCSI protocol was a no-brainer from the start. Others in the technology community, however, were not convinced, and even today the market is just now beginning to accept iSCSIas a viable alternative to running FC (Fibre Channel) as the transport layer in networked storage.
"We realized running storage protocols over IP was a good thing back in the end of 1997," says Julian Satran, a Distinguished Engineer at IBM's Haifa Research Lab and the lead co-author of the iSCSI protocol specification that was finally ratified by the Internet Engineering Task Force (IETF) in February. "We did measurements in 1998 and knew then it could achieve better or at least similar performance as Fibre Channel."
Unbeknownst to IBM, Cisco shared its belief in the technology. In early 1999, Cisco intern Costa Sapuntzakis was working under the esteemed technologist Andy Bechtolsheim in the Gigabit Switching Group. In doing research for what the Sapuntzakis, then 23, describes as a thinking project, the young technologist identified the SCSI protocol as a candidate to benefit from the Gigabit Ethernettechnology Cisco was developing. He drew up his draft of the iSCSI protocol and shared his idea of putting SCSI traffic onto an IP network with Bechtolsheim.
Bechtolsheim was impressed with the findings, and months later in a speech about Ethernet, he mentioned the idea of SCSI over IP in passing. IBM representatives were present and contacted Cisco, thus starting the work on iSCSI as a standardized protocol in late 1999.
Hewlett-Packard's Mallikarjun Chadalapaka joined IBM's Satran, Kalman Meth, and Efri Zeidner and Cisco's Sapuntzakis in 2000. Shortly thereafter, the engineers submitted an initial draft of the iSCSI protocol to the IETF, which was reluctant to even form a working group. The IETF believed the TCP/IP stack was too heavy to be used to transport storage data, explains Satran.
"Eventually it became obvious that using the TCP/IP stack was the right way to go because everything else is on it," Satran says.
"In four to five years, I think iSCSI will reach parity with FC in terms of technology," adds Chadalapaka, technical contributor of Networked Storage Architecture, Network Storage Solutions at Hewlett-Packard. "Will it overtake FC? That is a question for the market."
Chadalapaka points out that iSCSI has already made an impact and that the IP Storage Working Group's contribution will be significant. "From the beginning we were all convinced storage over Ethernet would happen," he says. "The economy still has to line up, but it will happen. It is happening already in a way."
(For profiles of the other nine 2003 InfoWorld Innovators, see Honoring the Innovators.)