-by Sean Fitzpatrick
The industry is buzzing about Fibre Channel over Ethernet (FCoE) and what its impact may or may not have on iSCSI and network convergence. Initially there were some skeptics in the press like the story from Howard Marks FCoE: The Latest Standard We Don’t Need. But now it seems the tides are turning as you can read about some of the activity on the FCoE website. Putting initial total cost of ownership (TCO) discussions aside; the technology has merit and worth looking at as a catalyst to network convergence and maybe some day replacing iSCSI.
FCoE is an emerging technology that isn’t new but rather revitalized from the work being done in the IEEE (Ethernet) and T11 / Fibre Channel Industry Association. Two improvements have been made to Ethernet include; 1) lossless Ethernet and 2) pause-base flow control. In the late 90’s these two enhancements were not available, hence why we ended up with iSCSI.
FCoE encapsulates Fibre Channel over an Ethernet transport combining the two technologies. In a real application environment, it makes good sense to leverage one common FCoE connection between the server and network enabling new networking options. The good news is FCoE will utilize the same common driver stacks, cabling and management applications you’re already used to dealing with today.
On the other side of the story, iSCSI requires the use of TCP/IP and as we all know has the tendency to drop data packets, unlike FC. Plus, you will take a performance through put hit because of TCP overhead. If you’re deploying iSCSI today you know it requires its own software and management tools separate from networking apps. Because of these characteristics, iSCSI is not suitable for enterprise data centers; rather it is relegated to non-SAN remote storage, stranded storage (i.e. DAS) or small-medium business segments.
It’s anyone’s guess how long it will take for the adoption to take hole, we believe the market will eventually decide the fate of iSCSI or success of FCoE, with a little help of course. If we’ve learned anything from past experience, two things need to happen in order for FCoE to become viable alternative for iSCSI; 1) near cost parity and 2) sustained growth adoption. The premise of converged networks is to help drive out some implementation costs, examples; less adapters, optics, cables will help reduce overall power consumption. This all translates into a much greener environment. HP believes for the foreseeable future (1-3yrs) these two protocols will co-exist, and in or around 2010-11 we should start to see one of them gain more ground over the other.
By comparison:
FCoE will scale, where iSCSI is stuck.
FCoE will be lossless, where iSCSI is not.
FCoE will leverage common management, where iSCSI will always be unique.
FCoE performance is suitable for enterprise data centers, where iSCSI is not.
HP likes the new possibilities FCoE offers to enable network convergence. FCoE has potential to scale in most IT environment from small-medium-enterprise utilizing the same physical Ethernet layer and common management applications in use today. It’s too early to tell, but iSCSI and FCoE will co-exist initially. Once near cost parity is reached…will there be much reason to continue iSCSI?
|
Printable version
