Startup seeks to move SANs off Fibre Channel
Craig Matsumoto
05/29/2000
Electronic Engineering Times
SAN JOSE, CALIF. - Startup Nishan Systems Inc. has unveiled a plan to unseat Fibre Channel as the protocol of choice for storage-area networks (SANs), and replace it with the native Internet Protocol (IP) connections prevalent in other parts of the network. But Nishan is not alone, as multiple efforts are under way to draw SANs away from Fibre Channel.
By essentially allowing storage devices to speak to Ethernet boxes, Nishan (San Jose, Calif.) hopes to turn SANs into a more commoditized market, able to use the routers available for Layer 2 Ethernet LANs.
In separate efforts, at least three proposals for non-Fibre Channel SAN standards have been filed with the Internet Engineering Task Force (IETF).
"There are a lot of people trying to make this happen," said Tom Harrington, director of strategic marketing for Gadzoox Networks Inc. (San Jose, Calif.). "It will happen, but you're going to see it in limited pockets at first."
Nishan developed its own protocol, "Storage over IP," for translating the block-based communication of SCSI into Ethernet frames. The company has trademarked the SoIP moniker and plans to submit the protocol to the Internet Engineering Task Force for consideration as a standard.
Details of SoIP won't be released for a few weeks, said Randy Fardal, vice president of marketing for Nishan . But the company has signed up multiple partners to use SoIP, and OEM products should start appearing within months, Fardal said.
In addition to having users of its protocol, Nishan hopes to build SAN switches itself with additional features. But if SoIP works as Nishan claims, any Layer 2 Gigabit Ethernet switch will do the job for a SAN.
A place for Fibre Channel
Nishan is not advocating the overthrow of Fibre Channel. In host bus adapters and device controllers, Fibre Channel still has a place, Fardal said, because applications often are written to take advantage of the block-based data transfers used by Fibre Channel and SCSI. But Nishan officials think SAN traffic could be handled using IP.
"The only reason we haven't been able to use the IP infrastructure so far is the end systems," Fardal said. "They have been dealing in blocks of data, for Fibre Channel."
SoIP consists of Nishan 's protocol to translate Fibre Channel's block-mode data into IP; APIs to attract developer interest; and a mechanism for chopping data into frames of about 1.5 kbytes each, to match Ethernet data.
But IP does have its handicaps. "It's not well suited to high-frequency, low-latency storage-to-storage or storage-to-system transfers," said Jay Kidd, vice president of product marketing for Fibre Channel fabric vendor Brocade Communications Systems Inc. (San Jose, Calif.).
Block-mode data transfers over IP are new, and that poses a barrier. "Although it sounds like a good idea, there's a lot to be learned over the next couple of years," said Harrington of Gadzoox. "Will it displace Fibre Channel in SANs? Absolutely not."
Security is one concern, as the concept of IP-based SANs suggests that data storage and backup might be carried out across the public networks. Performance also is an issue, as storage managers are accustomed to predictable, nearly deterministic response times, Harrington said.
Still, multiple IP-storage proposals are under consideration within the IETF. Gadzoox has submitted its own Fibre Channel-over-IP scheme. IBM Corp. and Cisco Systems Inc. have jointly submitted a proposal allowing TCP/IP traffic to mimic SCSI. And Adaptec Inc. has developed a protocol for storage over IP, targeting low-end LANs. This last area is the one most likely to see IP-based SANs come into play, Harrington said. "Outside of that application, it's going to be experimental stuff for a couple of years."
All these proposals are handicapped by originating from companies, Harrington said. "The probability that Nishan or any of these guys is going to have their proposal just passed by the IETF is remote, because the IETF isn't going to be too happy about a proprietary proposal that could lock in one vendor and lock out another."
Gadzoox's work in IP-based SANs is based on a specific case. Gadzoox officials see distance-not Fibre Channel itself-as the biggest obstacle facing SANs and the primary motivation for finding an alternative to Fibre Channel.
"The way SANs are being deployed today are as islands of storage," Harrington said. Gadzoox was looking for ways to get SANs connected at longer distances. Fibre Channel petered out at 80 kilometers, and ATM was too expensive, both in terms of the equipment and in the leasing fees to get bandwidth from carriers, Harrington said.
So like Nishan , Gadzoox decided to tap the ubiquity of IP networking and Ethernet. The company's yet-unnamed box will be marketed not as a SAN in itself, but as a means of connecting SANs to the network at greater distances than are possible under Fibre Channel.
Gadzoox also has a yet-to-be-named product, demonstrated at Networld+Interop, that can connect Fibre Channel, Ethernet and SCSI environments, Harrington said.
eetimes.com
May 29, 2000
******************
Have to check out gadzoox.
Jack |
