Fibre Channel and IP SANs: An independent view

Four possible scenarios regarding FC and IP SANs exist, and the stakes are huge for both end users and vendors.

By Tom Petrocelli

There has been considerable discussion and concern regarding the future of storage area networks (SANs). As is usual with new technology, filtering out the signals from the noise is often difficult. One of the more important discussions revolves around the effect that IP-based SANs will have on Fibre Channel.

Tom petrocelli
Technology Alignment Partners
Click here to enlarge image

To begin with, IP-based storage is a successful technology that has been around for years. Network- attached storage (NAS) uses IP as a transport protocol over Ethernet networks. NAS is widely used as a method of deploying storage in networks of all sizes and types, largely due to its ease of use and relatively low cost. The major limitation of NAS-if you can even call it a limitation-is that it only supports file I/O. For the vast majority of applications, this is fine. It is especially good for typical office applications such as word processing, and Internet applications, which tend to be file-oriented.

The most common method of handling networked block I/O, usually referred to as a SAN, is with Fibre Channel. As a protocol, Fibre Channel is well-suited to block storage. Unfortunately, it also suffers from compatibility problems (though not as much as it used to) and is relatively expensive. The difficulties of installing Fibre Channel and properly tuning a SAN are daunting for many IT professionals. The paucity of solid and affordable tools contributes to the difficulties of managing SAN environments. Add to that the lack of well-trained Fibre Channel technicians, and it is no mystery why SANs are not more widely used. The core technology is good, but the hurdles to success are still high.

SANs based on IP running over existing Ethernet infrastructure are an attractive idea. Ethernet, even Gigabit Ethernet, is inexpensive. Gigabit Ethernet network cards can be purchased for less than $100 retail. These network interface cards (NICs) are not server-class equipment, but it does give a sense of just how low the price point is. The familiarity with the base networking technology and the shear number of skilled technicians available also make IP storage attractive to IT departments. IP and Ethernet are inexpensive, available, and easy when compared to other networking technologies such as Fibre Channel.

Unfortunately, there are a few drawbacks to running SANs over IP networks. IP networks were never designed with block I/O in mind. The demands of storage, especially connection response time, are more stringent than most IP networks are designed to handle. Common network problems such as congestion, time-outs, packets arriving out of order, lost connections, and rerouting of connections begin to take on new meaning in a block I/O environment. Most applications that cannot tolerate the non-deterministic nature of IP networks such as databases have hidden behind a file system or server for just those reasons. Still, these are not insurmountable problems. Switches can be made to allow higher priority paths and virtual connections for storage protocols. Networks can be designed to isolate storage traffic from other traffic and cut down on congestion.

Another concern is compatibility, or interoperability-one of the biggest challenges facing Fibre Channel. Vendors of IP storage products contend that this will not be the case with their solutions. However, the Achilles' heel here is the tendency of storage standards to be loose, compared to networking protocols. It is not uncommon to see many "may" or "should" rather than "must" statements in storage standards documents. This helps create gray areas in the standards that can lead to incompatibility. Some of the new IP storage standards are like this. Again, however, this is not an insurmountable problem. As standards such as iSCSI make their way through standards organizations such as the Internet Engineering Task Force (IETF), there is a real opportunity to tighten up the standards and avoid major compatibility problems.

At the moment, Fibre Channel is still the only viable SAN technology for enterprise computing. This technology has achieved a reasonable amount of stability, and significant progress has been made on compatibility issues.

In addition, Fibre Channel is not content to stand still. Most Fibre Channel vendors have released 2Gbps products that, for the moment, give the speed advantage to Fibre Channel. This technology is also firmly entrenched in companies that need SANs most-large-scale enterprises. Considering the investment that has already been made in Fibre Channel, it is hard to see this slice of the market abandoning it anytime soon.

Still, Fibre Channel is costly and poses difficulties in deployment and maintenance. The benefits of SANs have therefore been relegated to a relatively few companies with deep pockets. For many other companies, however, it is simply not worth the time and effort to overcome the roadblocks.

What will be the result of all these changes in SANs? Will IP storage make SANs as common as databases? Or will IP storage fall on its face and leave Fibre Channel as the dominant SAN technology? Four possible scenarios concerning SANs need to be examined to answer these questions.

Scenario #1: IP SANs fail

The first possibility is the one predicted by those who cling to Fibre Channel as the only SAN technology. In this scenario, IP SANs do not become viable, fast enough.

A new generation of chips and software is required to allow IP SANs to be deployed in real-world applications. From adapters to switches and arrays, new ASICs are required to ensure that packets are delivered in order and on time and that they are not lost. For performance, host bus adapters/NICs will have to employ TCP/IP Offload Engines (TOEs), moving TCP/IP processing out of the operating system and onto the I/O card. While ASICs and software to do this exist, they are only in their first generation. Competing agendas among ASIC vendors could delay IP SAN adoption. This is precisely what happened in Fibre Channel. It took a long time before the ASIC vendors were able to effectively tackle compatibility problems because of disparate technologies.

The current business downturn could also hurt the adoption of IP SANs as well as Fibre Channel SANs. If R&D continues to be curtailed, or some of the nascent IP SAN companies cannot survive, IP SANs might die on the vine.

However, these are only minor threats. The biggest threat to IP SAN adoption is hype. Over-hyping technology in a down market is dangerous. The promises of SANs becoming inexpensive and easy to install/maintain need to be realized in a short amount of time or else the IT community will likely reject the technology. The expectation that hardware and maintenance costs will be roughly equal to those of Gigabit Ethernet must happen quickly. Otherwise, IT managers in midrange enterprises will reject it.

Scenario #2: IP SANs kill Fibre Channel

The opposite might also come true, and Fibre Channel will be killed off by IP SANs. This is a real possibility if certain events occur. The first factor is, of course, the success of IP SANs, which depends on the technology meeting its promises.

This scenario also depends on IP SANs addressing interoperability issues better, and more quickly, than does Fibre Channel. While many of the key problems of Fibre Channel compatibility have been dealt with, there are still some maddening issues, especially with switch compatibility.

The final nail in the coffin would occur if the cost of Fibre Channel products continues to be at current levels. Compared to Gigabit Ethernet prices, Fibre Channel costs are very high-more than two to three times as high in many cases. IP SANs would be very attractive to budget-constrained IT departments.

Of course, Fibre Channel would not die overnight. Enterprises that have already implemented Fibre Channel would continue to do so for many years to come. (There are still Token Ring and SNA installations in the LAN world despite the dominance of Ethernet.) Still, the market share of IP SANs would increase dramatically over time at the expense of Fibre Channel, with Fibre Channel becoming a legacy technology with little forward momentum.

Scenario #3: Both are successful

Another possibility is co-existence. This might happen if IP SANs fulfill their promises of ease-of-use and low cost but cannot match Fibre Channel's perfor mance. This would create the scenario where each falls into its own strata based on price-performance ratios. Fibre Channel would occupy the space where cost is not as important as reliability and performance, and IP SANs would rule in cost-sensitive environments. This détente scenario is attractive because it stakes out clear territory and guidelines for choosing each technology.

Scenario #4: Both find limited acceptance

What if the IT community greets IP SANs with a collective yawn? In this scenario, IT departments start to look seriously at SANs because of the perceived advantages of IP SANs. But they realize that the cost and effort of implementing SAN technology is still not in line with the rewards. At that point, SAN growth starts to slow, and the size of the market is determined by a handful of large companies. This is very possible if the IT community looks at the new technology that is supposed to bring SANs into their reach but still can't justify the expense and pain in light of questionable return on investment. For many companies where IT operations are mostly file-based with a few databases, this is a distinct possibility. In those situations, companies are likely to buy NAS instead of SAN. Or, they would consider products that converge, or combine, NAS and SAN technologies.

Consulting the crystal ball

The big "if" is compatibility. If interoperability in IP SANs proves to be much less of a problem than it was/is with Fibre Channel SANs, then IP-based SANs should dominate within the next few years. (My guess is within five years.)

Fibre Channel is excellent technology for SANs, but IP SANs have psychological and cost advantages. IP is better understood and inexpensive and has a huge installed base. Given the choice between having one type of technology used for multiple purposes or several specialized technologies for those same purposes, most IT managers will choose the former. Over time, it will keep costs down.

On the other hand, if compatibility problems stymie adoption of IP SANs, then damage will be done to the market as a whole. If this happens, SAN deployments will stall as the high end of the market begins to saturate and midrange and low-end companies choose not to adopt SANs because of perceived issues with cost, compatibility, and deployment difficulty. Fibre Channel simply cannot lead SAN adoption in those markets. And if it is believed that IP SANs are just more hype, then SAN adoption will reduce to a trickle.

Tom Petrocelli is president of Technology Alignment Partners, in Buffalo, NY, an executive and management consulting company focused on strategic planning, business plan development, product planning, and business development. Petrocelli can be contacted at petrocelli@adelphia.net.


This article was originally published on January 01, 2002