iSCSI provides a means to transport native SCSI commands over TCP/IP, allowing storage arrays to be shared over the Internet Protocol (IP) and storage area networks (SANs) to expand by utilizing Gigabit Ethernet (GbE) networks. Earlier this year, the Internet Engineering Task Force (IETF) approved the specification as a viable standard.
However, the protocol faces potential competition from — or, at the very least, potential friction with — the Fibre Channel protocol prevalent in most SANs today. And while some industry experts believe iSCSI will have an eye-opening effect on the storage industry as a whole, others feel that it will be nothing more than “business as usual.”
“I don’t see IP-based SANs and Fibre Channel (FC) SANs as competing with one another, but rather as complementing one another,” says Shaul Gal-Oz, CEO at SANRAD. He adds that even though IP-based SANs range in price from 2 to 6 cents per MB, a single path, non-redundant basic 2TB IP SAN for 20 servers costs approximately $40K.
“However, if you look at that same capacity, but need high performance, multi-pathing, and active/active fail-over, the cost for the entire package including host HBAs, management, connectivity, and high performance storage quickly rises to about $120K,” he continues. “With the higher end IP SAN, you basically have all the same features as a traditional FC SAN, and it can easily scale to 20 TB or higher.” Gal-Oz cites, as an example, the fact that a complete 8TB SAN would cost $160K for everything.
Gal-Oz believes that today's FC SANs address high performance, while IP-based SANs address simplicity, cost, and flexibility. One of the biggest misconceptions about IP-based SANs, according to Gal-Oz, is that they don’t have the same features as FC SANs. However, the reality is that IP SANs have almost completely closed the gap in terms of offering the same capabilities.
Diamond Lauffin, Nexsan's senior executive vice president, does not see IP-based SAN installations as having a great effect on the market in general, but he does believe the presence of such options could potentially be disruptive to the process of the purchase decisions made by corporate IT departments.
“In my experience, I see corporate IT investigating more aggressively the question of what can be done with IP — whether it is NAS or iSCSI,” he says. “The anticipation of 10 Gigabit Ethernet becoming available is opening up the minds of forward-looking IT engineers to explore the possibilities. At 100 Mbps Ethernet it was a dream. Even if it could have been practically accomplished, there was no interest in driving it.
"Now, at 1 Gigabit Ethernet, we have a different horizon and a level of practicality and acceptance. As we look toward 10 GbE, the landscape of what will be possible with IP changes greatly,” he concludes.
The Many Benefits of IP Storage
Some folks in the corporate IT world feel that there are certain benefits of moving from DAS to SAN, and that these benefits apply to IP-based networks as well. Gal-Oz cites some of these benefits as the ability to share storage among many servers, thus better utilizing storage investments.
He also says that high-speed block level support for server clusters is not possible with DAS and NAS, and that server-less, LAN-free backup is faster, easier, and more reliable. Gal-Oz believes that centralized security and data protection are also major benefits, as is the fact that it's much easier to upgrade servers, add servers, add storage, or increase storage capacity per server.
Lauffin believes the benefits of IP SANs will evolve to where they will be nearly identical those of Fibre Channel SANs. “We see the feature sets of IP SANs encroaching steadily on the feature sets of traditional Fibre SANs. The longer that IP SAN technology has to mature, grow momentum, and attain acceptance, we are going to see more support poured into the funding of companies behind this technology,” says Lauffin.
He believes that as this happens we will see continued “Feature Compression” between the two technologies. “The IP companies already have a roadmap, they already have a blueprint — they are copying Fibre SANs — and we will see that the longer these companies stay in business and the longer they grow, the closer they will come to duplicating Fibre SANs across the Ethernet,” he concludes.
Great Potential for Mass Migration from DAS to IP SANs
Another major issue concerning IP-based SANs is whether or not we will start seeing DAS-entrenched companies puchase IP SANs en masse while reserving Fibre Channel-based solutions for high-end niches. Gal-Oz expects this to occur and says the reason is that there are a large number of users and applications out there that never migrated to a SAN due to the costs of server-attached Fibre Channel holding steady in the neighborhood of $2K to $7K per server, and customers still must add FC switching port costs, complexity, and lack of inoperable standards to the mix.
“The vast majority of our customers are migrating from DAS to SAN and would not choose FC – not because FC wouldn't do the job, but because they are more comfortable with iSCSI over TCP/IP or Ethernet, since this they understand, and the costs of deployment and maintenance are far lower,” he says.
“While SANs in general will become much more widespread as costs come down and manageability improves, says Lauffin, we will also see DAS storage systems evolving to contain embedded intelligence that will support features of existing SANs. However, he adds, Fibre will remain the weapon of choice for certain high-end applications.
As most people know, creating a snapshot or mirroring a volume for a host is a relatively easy task. However, ensuring that the data content stays consistent requires close interaction with a variety of host applications including databases, e-mail systems, and volume managers. As such, this is one area where Fibre is expected to hold court for the foreseeable future.
The Key to Market Success for IP SANs
So, will the market share for the new IP SANs vendors depend on how well they cover the entire storage puzzle? Absolutely, says Gal-Oz. “Customers expect and need all the same functionality within an IP SAN that is found in a FC SAN,” he says. “A SAN is a SAN and must have provisions for high availability, disaster recover, security, and failover.”.
Gal-Oz says that IP SANs deliver these today, including snapshot, path-failover, and 100 percent synchronous mirroring of individual storage systems. “Even advanced features like asynchronous remote mirroring and DR are right around the corner for IP SANs.” And synchronous remote mirroring over IP exists today (using iSCSI and new operating system features at the host layer) and will soon be available from the network layer, he continues.
Lauffin says that what he has seen clearly over the last six to nine months is the concept of IP storage transcending all demographics. He says he sees it in the vertical OEM markets — such as digital video security, medical imaging, and print graphics — and across the board. “What in many instances was traditionally a server-to-storage device DAS relationship is today, with the affordability of enterprise class storage systems, an entirely different ballgame,” says Lauffin.
Lauffin points out that these systems offer vast capacity in comparison to past RAID systems. “The capacity ratio of available storage is now so far beyond the original DAS concept that the desire to make that storage accessible to multiple end users on similar type systems has become a regular request, “ he says. “As these requests increase, the ability to access that data seamlessly over something such as an IP network is very attractive,” he continues.
Paving the Road for the Future
It took more than three years for iSCSI to gain the IETF's seal of approval, and some industry experts feel that with its ratification, 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.
While the debate continues in the storage arena, we will continue the “Storage Focus: The Outlook for IP SANs” discussion in Part 2 of this series. The second column will appear later this month and will answer such questions as:
- Should organizations be looking at pure IP-based storage as a storage solution? If so, what type of organizations would most benefit from IP-based SANs?
- IP-based storage products and applications are becoming increasingly important components of the modern storage networking world. However, IP networks suffer from latency, jitter, lost packets, and other common networking artifacts because they transverse wide areas and comprise switches, routers, and complex protocols. What can IT managers who are considering a pure IP-based SANS solution expect or do to minimize these issues?
- One way or another, enterprise storage is going to be networked. The question is how: Fibre Channel (FC) or Internet Protocol (IP), or both?
- Are more and more companies beginning to employ iSCSI at the edge for less critical or performance-intensive storage, and then using it or other IP protocols to connect to central FC storage area networks (SANs) across the wide area?