Download the authoritative guide: Enterprise Data Storage 2018: Optimizing Your Storage Infrastructure
The days may be numbered for the dominant data storage interconnect technology.
I have been tracking the prices of 4Gb and 8Gb Fibre Channel (FC) and 10Gb Ethernet (10GbE) for more than a year, and just recently I have started to see what I have been expecting: Dramatic price drops in 10 GbE.
The price drops started with NIC pricing, as that is the easiest and lowest-cost component to develop, store and ship. I believe that this price drop, combined with FCoE and other market forces, means the slow death of Fibre Channel has begun. It will not disappear today, in 2010 or even 2011, but these forces will change the storage horizon relatively quickly, given the potentially huge cost savings. Let's explore why I think this change is coming.
FibreOn Never Took Off
In the early part of this decade, there was a movement called FibreOn. The idea was to increase the market for Fibre Channel by putting Fibre Channel chipsets on higher-end motherboards. The plan failed miserably for reasons that I think are pretty clear.
First, SATA drives became denser than Fibre Channel drives, their reliability increased, and their cost dropped compared to Fibre Channel drives.
Secondly, Fibre Channel chipsets were expensive, and vendors such as Dell (NASDAQ: DELL) and HP/Compaq (NYSE: HPQ) had a chicken-and-egg sales and marketing scenario. Since margins on PCs were slim, building and designing a board with Fibre Channel chips and plug-in SFPs would be a relatively high cost, so there had to be a market for them, which never materialized, in part because there wasn't much in the way of products. So FibreOn failed, and during this time vendors started to add 1Gb Ethernet to their boards.
Today, the cost of 1GbE on PC boards is likely less than $10, and definitely less than $20. You can buy 1GbE NICs for less than $10 from a number of retailers. Prices for these products were much higher six months ago, and even higher a year ago. So Fibre Channel failed to gain much PC market share and was relegated to peripheral storage, and that storage saw a significant drop in market potential. As a result, Fibre Channel connectivity pricing stayed relatively high, which further limited the market.
10GbE Pricing in Freefall
I have been tracking pricing for more than a year using the same Web site of a major online reseller. These might not be the best prices, but they are surely competitive. Here is what I found for Fibre Channel pricing back in December 2007, when I started thinking about this trend:
I looked at the same FC pricing today and found that there's not much difference in price compared to 2007. This was not a huge surprise based on what I see in the market.
Looking at the same data for 10 GbE shows an astonishing drop in prices over the same period, from December 2007 to today:
Clearly there have been some significant drops in 10GbE pricing recently, well above 80 percent. I expect that the drop is just starting, as I recall a similar trend when 1GbE hit, and soon thereafter the cost of the chipsets dropped to make them cost-effective enough for motherboards.
Fibre Channel still has a more than two-to-one price advantage in the cost of switch ports, but I don't think that advantage will hold out for very long, based on past trends and the usual lag in the drop of switch pricing compared to NIC pricing. It takes time to build switch boards, so they reflect the cost of chips during the past. I expect that by mid-summer, 10GbE switch ports will be competitive with Fibre Channel on both a cost per Gb/sec and a cost per port basis. What that means is that for the first time in a very long time, networking performance will exceed the performance of the standard enterprise storage connection. Some might argue that InfiniBand already exceeds the performance of Fibre Channel, but from what I see, it's just not used in enterprise storage.
Network Convergence Dead Ahead
This dramatic pricing trend has a number of broad implications for the next three to five years.
- The convergence of storage and communications networking: For as long as I have been working on computers, these have been different networks, and that is about to change, in my opinion, based on the sheer logic of price. Why else would the industry Fibre Channel leaders (Brocade, Emulex and QLogic) all have 10GbE products and plans if the winds of change weren't blowing?
- The use of protocols over Ethernet other than TCP/IP: As I and many others have discussed, FCoE (Fibre Channel over Ethernet) is coming, and combined with some of the Ethernet enhancements, it is going to change data center connectivity starting in 2010 or 2011. This will not be a wholesale change, given the investment in Fibre Channel over the last decade, but just like other changes, such as disk technology, it will happen.
- SAN file systems and large shared file systems might be combined with pNFS, and lower-cost Ethernet communications might change the way file systems communicate, and they might even change the SAN/NAS ratio and usage model. Today NAS is usually not used where applications are sensitive to latency, and SAN shared file systems are not used for general-purpose bulk sharing of information. This will likely change and these technologies will begin to morph.
Falling Ethernet costs, improved reliability and high availability, and the rise of new protocols like FCoE will change the current data center — and will change everything, I believe, from home computers to large businesses to HPC. Technology may drive the market, but unless the technology gains enough market acceptance and becomes a commodity, it will never be anything but a niche. We have seen this time and time again in networking, from FDDI back in the late 1980s to HiPPI in the early 1990s, and now to Fibre Channel. Rightly or wrongly, the market is driven by commodity technologies. If it is not going to fit within the home environment, PC, TV, gaming, music and so on, then it will not stick around forever.
Fibre Channel was a niche product that nonetheless achieved market dominance for more than 13 years, and it will continue to dominate for a bit longer. The end is near, but it was a hell of a run, and it fulfilled a need that took commodity technologies years to fill. There are other niche markets and technologies in the networking area such as InfiniBand, which right now is at 40 Gb/sec. It might surprise you to learn that the 40 Gb/sec Ethernet standard is not far from ratification. Things are changing fast, and even the Fibre Channel vendors have embraced the change that is coming. The 10-gig Ethernet price drops of the last few months have made that change inevitable.
Henry Newman, a regular Enterprise Storage Forum contributor, is an industry consultant with 28 years experience in high-performance computing and storage.
See more articles by Henry Newman.
Follow Enterprise Storage Forum on Twitter