NEW YORK – Trying to nail down a consistent sentiment about iSCSI here at the Storage Decisions 2004 conference in New York City is like trying to type without a keyboard – near impossible.
Depending on who you ask about iSCSI (and possibly at what time of day you ask them), answers range from a 20-minute plug-and-play solution to a technology that is years away from implementation.
At this point, there are essentially three camps of users when it comes to the iSCSI discussion. In the first camp are extreme advocates of the technology. Brave souls that have done the testing, installed iSCSI and are using it in a production environment.
Camp two are the users that have faith in iSCSI but are only using it in quality assurance (QA) environments or doing extensive research about the technology. The third camp houses skeptics that are still waiting for the technology to get up and take its first steps and mutter first words.The champions of iSCSI
Michael Davies, chief implementation officer at Sawtel Inc. a satellite communications provider, Thomas Reynolds, senior executive director, information systems and technology at Idenix Pharmaceuticals and Ken Walters, senior director, enterprise platforms for PBS have all taken the leap into the world of iSCSI. All report success and similar reasons for implementing – centralization and major cost savings – without sacrificing the functionality of a Fibre Channel storage area network (FC SAN).
"I've had a Fibre Channel SAN for about four years now," said Walters. "About a year-and-a-half-ago I realized I had only attached the heavy-lifting [servers] to the SAN and the majority of the servers were still direct-attached (DAS). I desperately wanted to go from DAS to a centralized format. With iSCSI I was able to bring the stranded servers [to the network] for much less [money]."
Walters performed rigorous tests on iSCSI in a QA environment and has been rolling out severs on a private iSCSI network since testing ended in 2002.
Davies was also lured to iSCSI by the cost savings. He says he is paying 1/6th of the costs of a FC SAN without losing many of the features – including performance – which has been a knock on iSCSI for some time.
"iSCSI performed really well. Since I am in the satellite business the only performance hit was latency caused by distance," said Davies. "I'm noticing 90-100 mb/sec transfer rates."
Walters also compared the performance of iSCSI and network-attached storage (NAS). Where iSCSI won out in his mind was the fact that iSCSI has block I/O, whereas NAS is only good for file level I/O. Neither Reynolds nor Davies have found it necessary to employ TCP offload engine (TOE) cards, because they say it would not have an impact on performance. Walters is only using TOE cards in servers that he feels need an extra boost.
Where iSCSI really surprised users was in the deployment phase.
"It took me about 20 minutes to set up the cables, NICs, etc." said Reynolds, whose environment is predominantly Windows-based. "The only surprises [in implementing iSCSI] I've had have been good surprises."Close, but not quite there yet
Rick Husley, manager midrange engineering and support for Southwest Airlines is excited about iSCSI. "We are looking at iSCSI going forward," said Husley. "We have run tests on iSCSI and FC and found they run neck-and-neck. When you look at cost per port and complexity you have to ask yourself 'Do I really need to build a big FC SAN?' Probably not."
Right now Husley has iSCSI only in that testing environment. Similarly, Robert Stevenson, technology strategist from Nielsen Media Research is kicking the iSCSI tires. Stevenson says he has iSCSI in his QA environment and is planning on beating it up in the next few months with a ton of data. Stevenson says he has been hesitant to implement iSCSI because he feels the technology is still a bit immature.
A couple of years away
Kris Knutson, director of infrastructure services for Carfax isn't totally against iSCSI. But for him iSCSI is still at least "two years away." Knutson, who has had a Fibre Channel SAN for a few years, is headed towards deploying NAS. He says in about two years he'll reassess where iSCSI is and determine if he'll deploy a solution.
Even more skeptical about iSCSI is SearchStorage.com's Storage Networking A-Z expert and founder of Building Storage Inc., Marc Farley. Farley cites a couple of reasons why he is a bit wary of iSCSI actually making a big impact anytime soon.
"iSCSI has been breaking my heart," said Farley. "The problem is many users are too conservative to implement iSCSI; and are too concerned about losing data."
One aspect of iSCSI that Farley is adamant about is how TOE cards will not necessarily speed up your iSCSI environment. "TOE cards drive me nuts. You don't need them. If your servers have a processor of 2.4 gigahertz or more you don't need them," said Farley.
Farley also says there is no good financial reason for vendors to offer good iSCSI solutions. "iSCSI will reduce everybody's margins. Why would a vendor sell iSCSI when they can sell FC for 2 or 3 times more? I don't think iSCSI is imminent. I just don't see a wave of products coming."
FOR MORE INFORMATION:
Get Marc Farley's slide presentation