Storage over IP: What's the protocol?
Storage over IP: What's the protocol?
July/Aug Edition, 2007: At first glance, the confrontation between fibre channel protocol and IP-based storage protocol iSCSI could be seen as just another turf war served up by the IT industry - Microsoft vs. Sun vs. open source, J2EE vs. .Net, Macintosh vs. personal computer. Darren Baguley investigates if the decision of IP Vs FC really just an emotional decision, like Holden vs. Ford, and if there really are differences between the two protocols.
While some proponents might be almost religious about the choice, the reality is that both protocols already co-exist in large organisations. Fibre Channel over IP is the best way to connect SANs over distances of greater than 100 kilometres and has been for a long time. As a result, it is widely used for disaster recovery. But when it comes to localised storage networks, until recently the high price of fibre channel restricted it to the very top tier of companies.
And even though iSCSI has been around as a mature protocol for at least two years, in terms of customer comfort in deploying it’s still new and struggling to gain mainstream acceptance. Part of the problem is that it took nearly two and a half years before the iSCSI protocol standard was ratified. In the interim, a lot of FUD (Fear, Uncertainty, and Doubt) was put into the market - mainly by vendors who were a long way from launching product.
Since then, the gap has closed between the two protocols. With vendors such as EMC and Network Appliance shipping SANs and NASs with both fibre channel and IP connectors, it is rapidly becoming more than anything else a case of ‘horses for courses,’ or what storage protocol is most appropriate for a particular application?’
IDC’s associate vice president for storage Asia Pacific, Graham Penn, says that in 2006 Australian organisations spent $6.6 million on iSCSI-based hardware. He’s predicting that figure to nearly double to $12 million in 2007. The majority of this growth is coming from small to medium sized companies that have yet to deploy a fibre channel SAN.
“When iSCSI first appeared, fibre channel was not yet mature technology. Hardware such as director-class switches and host bus adapters (HBAs), were priced at a premium, as were the skills required to setup and maintain the network. As fibre channel technology has matured and become commoditised, the price gap has closed,” says Penn.
“There are now enough trained people at vendor, reseller or company level so for small companies that want to go fibre channel, cost or lack of skills shouldn’t be an issue. But with iSCSI, companies have got IP skills in house so it’s simpler, plug the device in and Windows can see it. For this reason, iSCSI’s main applicability will be to entry level companies, particularly to those that have never deployed a fibre channel SAN.”
But for companies that have already deployed fibre channel, Penn believes there’s probably no particular reason to go to iSCSI. “There’s a minimal cost advantage to iSCSI, but it’s disruptive to what’s already there so there’s no point (taking out what’s already there)” he says.
NetApp director of marketing and alliances ANZ, Mark Heers, echoes Penn. “The gap has closed to the point where fibre channel is about 10 percent faster than iSCSI, but IP-based storage networks are a bit cheaper. So we’re finding iSCSI is popular in the mid-tier where companies haven’t made the step to SAN at all and they can avoid the whole adoption of and the costs associated with the original rollout of fibre channel,” he says. “Although the cost of fibre channel has come down, the bottom line is that it has natural costs over iSCSI. You need an HBA for each server, two to allow for redundancy, whereas servers have already got Ethernet ports onboard.”
EMC’s product marketing manager, Clive Gold, is also seeing a lot of activity around iSCSI among small to medium businesses (SMBs), but a lot of large enterprises that already have fibre channel networks are also deploying fibre channel. “It all comes down to which applications are suited to IP storage. Generally where you have lots of small packets of information being passed around, applications such as email and some transactional systems like databases, IP works because the handshaking is a lot more efficient.
“But for applications that have large volumes of data, fibre just blows IP away,” continues Gold. “Systems such as decision support systems, market research, business intelligence and large transaction systems such as those used by banks and Telco’s will always use fibre channel because of the volume of data.”
According to Heers, much of this ‘horses for courses’ approach has been driven in the market by Microsoft’s support of the iSCSI protocol. “Microsoft have got fully behind Exchange on iSCSI, it proactively talks about it and has produced patches to backdate its operating systems for iSCSI,” he says. “A lot of top-tier companies may have a fibre channel SAN but use iSCSI for their Exchange servers.”
HP Storage marketing spokesperson, Mark Nielson, adds that many companies are adding second tier servers into their SAN that normally would be hard to cost-justify connecting to the SAN using fibre channel. “Because of the cost of HBAs and the like, there is a higher cost to connect to SAN environment than an IP network. iSCSI is a viable technology in the data centre when organisations want SAN connectivity but don’t want the expense of fibre channel.”
Setup and management is also an area where iSCSI is making a big difference. According to Gold, EMC is about to release a programme that will get an IP SAN up and running in 15 minutes. “It’s power to production in 15 mins. The user goes through a wizard with five screens and 20 questions, and an IP SAN is up and running in 15 mins. HP has a similar approach. “We have an entry level IP storage box that can be up and running in 10 mouse clicks. Plug in, browse and set up in 10 mouse clicks via a nice automated GUI.”
In all cases, the key to a successful deployment of an IP-based storage system is having a good Ethernet network infrastructure in place. At the moment, for any organisation wanting decent performance, gigabit Ethernet is a base requirement and as 10 gigabit Ethernet becomes mainstream in a couple of years it will become the standard. “As 10 gigabit Ethernet becomes affordable, IP will also have the raw speed of fibre channel.” says Gold.
As a counterpoint, Brocade’s country manager ANZ, Graham Schultz, is not as sanguine about the future for iSCSI. “As the cost of fibre channel connectivity reduces, so the delta between the benefits of iSCSI and fibre channel reduces and customers are questioning if it’s worth pursuing the iSCSI route,” he says. The majority of customers looking at deploying iSCSI are mainly doing so for low cost connectivity to the fibre channel SAN for low-end Windows servers.