In Part 1 of this Storage Focus series on IP SANs, we discussed how with iSCSI recently gaining the IETF's seal of approval, some industry experts are predicting many startups and large storage vendors will redouble their efforts to develop storage arrays, host bus adapters, TCP offload accelerator cards, and management software products that support the standard. Let's continue our look at IP-based SANs and resume the debate over whether Fibre Channel (FC) SANs or IP SANs will ultimately prevail over the other.
It has been said that two of the major benefits of IP-based SANs (versus FC-based SANs) are gains in operational efficiency and reduced hardware component pricing. And even though there are still some that may disagree, many industry experts feel that IP-based storage solutions cover a wide range of markets, with a sweet spot potentially in mid-range applications.
There are also those who say that organizations should not be thinking about IP-based storage as an overall storage solution. But the main concern of IT departments remains whether or not the technology can be beneficial, and if so, what types of organizations will gain the most from using it?
Diamond Lauffin, senior executive vice president with Nexsan, says it's difficult to pinpoint what type of organization would benefit the most form this technology; however, he does feel that data center environments in which Fibre is not already installed and there isn’t a budget to install Fibre Channel would most likely find IP-based SANs to be an attractive option.
“IP-based storage is clearly going to take a position and is going to answer the needs of many organizations,” says Lauffin. These organizations would consist of companies that need flexibility in areas where they would not consider putting in a Fibre SAN for a variety of reasons, even if the cost differential came down.
For example, Lauffin says if a company has an existing Ethernet network that can support the current IP-based solutions and eventually 10-Gig E, there would be a great deal of attractiveness in deploying an IP solution as opposed to going through FC – especially at a departmental level and at a branch level. “This is where I see a significant level of interest and deployment,” says Lauffin.
Mid-Range Apps Better Suited for iSCSI?
Shaul Gal-Oz, CEO at SUNRAD, has found that mid-range applications are better suited for iSCSI. “Some lower-end customers (and small departments with low performance requirements) can still use NAS, while higher-end applications such as rendering, OLTP, and scientific simulations continue to use FC.” However, file servers, Web servers, messaging/email servers, data streaming systems, imaging systems, mid-range video streaming, print servers, etc. will operate very well within an IP-based SAN environment, says Gal-Oz.
IP-based storage products are becoming increasingly important components of the modern storage networking world. In Camp 1 are those who feel that because IP-based storage products comprise switches, routers, complex protocols, and traverse wide areas, the technology will inevitably suffer from latency, jitter, lost packets, and other common networking ailments. On the other side of the fence are those in Camp 2 – these are the folks who believe that the above are simply quality of service issues and that the application requirements for performance are of more importance.
“First off,” says Gal-Oz, “iSCSI is designed to deal with lost packets and slow performance within poorly designed solutions, and the session layer addresses these issues.” For example, he says, “we can back up and recover a server to and from storage that is located in another state or even another country,” although he did admit that this would obviously be slow if the user elected to perform this task over the Internet.
But he says the important thing about all of this is that it works, and his company has done it. SUNRAD has also successfully conducted data migration and replication over a line-of-sight infrared connection between buildings three miles apart and has built wireless SANs. Gal-Oz advises any IT department considering an IP-based SAN environment to work with an iSCSI product supplier that not only understands storage, but that is also extremely competent in TCP/IP networking.
As with any given technology, early on there are bugs, and over time those bugs are worked out and significant improvements are made. IP SANs are no different. Lauffin believes the majority of the issues that people speak about today concerning IP-based networks have all but gone away. “There are safeguards in place in the better IP systems, and these systems eliminate those concerns,” he says.
The Road to Networked Storage
One way or another, enterprise storage is going to be networked. Will it be through FC? Will it be through IP? Or, will it be some combination of the two? Lauffin says he sees both technologies as having a substantial presence within the IT industry. “There's going to be a place for FC – because there are certain protocols in the FC world that are still more advanced than the IP world,” he says.
Gal-Oz agrees but also says that because user preferences and applications are so diverse, this is what will dictate the choice of FC or IP. He also believes that many organizations will use IP to extend their FC SANs. “These users have already invested in a robust FC SAN solution and would like to leverage this investment by connecting more servers,” he says.
Gal-Oz goes on to say that FC connectivity for smaller servers may be too expensive or not possible due to lack of FC infrastructure. The solution, he says, “is to use an intelligent iSCSI with a FC SAN switch that includes security, multi-pathing, and volume management.”
Gal-Oz believes this will allow the FC SANs to ‘fan-out’ and provide simple and inexpensive connectivity for hundreds of small servers. In addition, he says that the connectivity and management costs can range from ‘zero’ dollars to under $1,000 – depending on the performance requirements or need for high availability.
Deploying iSCSI for Misson-Critical Applications
It seems that more and more companies are deploying iSCSI at the edge for less critical or performance-intensive storage, but are they using it for mission-critical applications? “With the adoption of any new form of technology, most companies that are attracted to the concept and believe in the concept – in this case, IP storage and IP SANs – will first deploy it in a non-mission-critical application. As they become more comfortable with the technology and the technology proves itself, we will begin to see it deployed in more mission-critical applications,” Lauffin says.
Gal-Oz firmly believes the idea is catching on and points out that many of the traditional enterprise-class FC SAN storage suppliers are making iSCSI available on their high-end storage products. “I think there will be a huge drive to connect as many servers as possible to the data center, and many of those servers will be connected using iSCSI,” he says.
He cites campus environments (e.g. large universities, hospitals, and state and local governments) as prime examples because these institutions have large LANs spread across hundreds of acres and spanning many individual buildings. “Campus environments tend to have a central IT group responsible for storage services from the data center, and only 10 to 25 percent of their servers are within the data center – the rest are scattered around the campus with no FC connection to the data center.” However, he did say that they are all connected via a LAN, such as a Sonnet ring, and this is what provides the connectivity layer for iSCSI over TCP/IP.
Gal-Oz feels a common application for the data center FC may be to provide backup and disaster recovery services to all the departmental servers on a campus by using the existing backup software to drive backup files over the Sonnet ring to an intelligent iSCSI multi-protocol switch, which in turn writes the files for the FC-SAN.
In addition, he also believes that another application might handle the vaulting of critical or sensitive data that cannot be resident on unsecured servers or that might provide storage for the ever-expanding email databases. “Eventually, instead of only 10 percent of servers on a campus connecting to the SAN, 90 percent of the servers will be connected using a combination of iSCSI, FC, and intelligent iSCSI multi-protocol switches,” he concludes.
Whether the future is FC, iSCSI, or more likely a combination of the two, the common requirements for all storage customers will remain the same – to reduce costs, minimize overhead, streamline administration, and make storage simply an additional component in their overall network strategies.