Data Center

The Business Case for Changing Technology

by on ‎04-29-2011 12:21 PM (396 Views)

Jon Oltsik at ESG wrote a post for Network World titled "Is Ethernet Storage On the Horizon".  He provides survey data on customer plans to deploy FCoE storage.  His post prompted me to comment.  I think there is another way to view the survey results, And I believe there are only two primary use cases for making any strategic change in technology such as elimination of native Fibre Channel and substituting FCoE.

I've taken the liberty to snip out the relevant statistics he quoted:

  • 42% Yes, but no specific timeline at this point
  • 32% Yes, we are doing this today
  • 5% Yes, over the next 12-24 months
  • 21% No, we will continue to run separate networks for data and storage in the data center

A former boss told me there are only two priorities in life, what you are doing now, and what you are ignoring.  Another way to interpret the results is that nearly two thirds (63%) of respondents don't have a compelling reason to deploy FCoE today. That would be the sum of the 42% who have no specific timeline and the 21% who said "No".  In my years of observing human nature, when someone says they don't have a specific timeline, then that's a polite way of saying its in the "ignore" stack on their desk.   

I think a new technology that competes with an existing one can only get significant market traction if it satisfies at least one of these two use cases.

(1) If the new technology can do the same thing the incumbent can at 1/2 the cost or less. Without that economic leverage to offset the risk inherent in the “new”, rational folks have no compelling reason to change technology. Said differently, “If it ain't broke, don’t fix it”.

(2) If the new technology can do something the old can’t and that something HAS to be done. My example of this is Fibre Channel vs. DAS in the early part of the last decade. Storage HAD to grow but DAS limits on distance, speed and connections prevented it. Storage growth is an unstoppable force creating the compelling reason to change technology (from dumb copper wire to optical switched layer 2 network). The storage industry adopted Fibre Channel, and all the “new technology” risks inherent, because there was no choice. Said differently “Sometimes ya gotta do what ya gotta do.”

There is a third use case where a new technology can get traction, and that is when it serves an unserved market with good enough capabilities.  This situation is amply covered by The Innovator's Dilemma and The Innovator's Solution, by Clayton Christensen, et al. I don't believe this is applicable to the FCoE vs. Fibre Channel decision. FCoE does not address an unserved market with "good enough" technology.  I think any unserved market was addressed last decade when iSCSI arrived.  Said differently, I don't know what unserved market FCoE addresses. But, I'm always open to education.

Thinking about FCoE, I don't think it does something storage administrators must have done that Fibre Channel can’t do. For that reason it fails to meet the criteria of use case (2).

Use case (1) is more interesting.  It may happen as 10 GE LOM arrives and becomes ubiquitous in the next generation of servers. The 1st connection to the storage network is expensive. This is compounded for server connections due to IO requirements of servers vs. IO performance of array ports. As a rough rule of thumb, there are an order of magnitude more server connections than storage connections since servers require about 1/10 the IO the array port is capable of. Of course, your mileage may vary.  Applying this rule of thumb to 10 GE LOM, 10 GE LOM is “free" and would change the SAN cost equation in favor of FCoE. The question is, is it enough of a saving to satisfy the criteria of Use Case (1)?  Time will tell.

Since FCoE preserves FC technology, knowledge and skills, storage administrators won't have to learn a new technology, so it’s conceivable that classic FC networks can evolve into FCoE networks over time.  This transition would be driven by cost.

Are there any compelling storage requirements that Fibre Channel can't provide but FCoE can?  If you have some, please share.  I'd appreciate your insights.