Just When You Thought It Was Safe To Go In The Water Again!

April 4, 2008 – 4:12 pm

In the shark infested waters where I/O and networking debates often rage, the Fibre Channel vs. iSCSI, or, is that iSCSI vs. Fibre Channel debates continue which is about as surprising as an ice berg melting because it floated into warmer water or hot air in the tropics.

Here’s a link to an article at Processor.com by Kurt Marko “iSCSI vs. Fibre Channel: A Cost Comparison iSCSI Targets the Low-End SAN, But Are The Cost Advantages Worth The Performance Trade-offs?” that looks at a recent iSCSI justification report and some additional commentary about apples to oranges comparisons by me.

Here’s the thing, no one in their right mind would try to refute that iSCSI at 1GbE levering built-in server NICs and standard Ethernet switches and operating system supplied path managers is cheaper than say 4Gb Fibre Channel or even legacy 1Gb and 2Gb Fibre Channel. However that’s hardly an apple to apples comparison.

A more interesting comparison is for example 10GbE iSCSI compared to 1GbE iSCSI (again not a fair comparison), or, look at for example the new solution from HP and Qlogic that for about $8,200 USD, you get a 8Gb FC switch with a bunch of ports for expansion, four (4) PCIe 8Gb FC adapters plus cables plus transceiver optics which while not as cheap as 1GbE ports built into a server or an off the shelf Ethernet switch, is a far cry from the usual apples to oranges no cost Ethernet NICs vs. $1,500 FC adapters and high price FC director ports.

To be fair, put this into comparison with 10GbE adapters (and probably not a real apples to apples comparison at that) which on CDW go from about $600 USD (without no transceivers) to $1,100 to $1,500 for single port with transceivers or about $2,500 to $3,000 or more for dual or multi-port.

So the usual counter argument to trying to make a more apples to apples comparison is that iSCSI deployments do not need the performance of 10GbE or 8GbE Fibre Channel which is very valid, however then a comparison should be iSCSI vs. NAS.

Here’s the bottom line, I like iSCSI for its target markets and see lots of huge upside and growth opportunity just like I see a continued place for Fibre Channel and moving forward FCoE leveraging Ethernet as the common denominator (at least for now) as well as NAS for data sharing and SAS for small deployments requiring shared storage (assuming a shared SAS array that is).

Iím a fan of using the right technology or tool for the task at hand and if that gets me in trouble with the iSCSI purist who wants everything on iSCSI, well, too bad, so be it. Likewise, if the FC police are not happy that Iím not ready and willing to squash out the evil iSCSI, well, too bad, get over it, same with NAS, InfiniBand and SAS and that’s not to mean I donít take a side or preference, rather, applied to the right task at hand, Iím a huge fan of these and other technologies and hence the discussion about apples to apples comparisons and applicability.