Would we have done things differently had we known about UCS?

Readers who have followed me as an independent writer, then as a blogger at Equallogic and later at Dell before coming to 3PAR might be familiar with my thoughts on the wisdom of keeping data networks and storage networks separate.   This led me to take a position opposing FCoE and some sparring with Chuck Hollis over the matter. 

Last year I gave up on the EqualLogic-born idea that iSCSI could become the defacto SAN network after HP acquired Lefthand. It's possible that native iSCSI could become the defacto way to connect storage over 10g Ethernet networks, but I've lost faith. FCoE is going to be the leading 10g SAN technology.

That said, I maintained my opinion that isolating LAN and SAN networks for fault determination and change management was prefereable.  In other words, run FCoE on a separate network from LAN and IP traffic.  However, after Cisco's announcement yesterday I'm starting to see the potential advantages of a superior virtual network infrastructure for everything. If excellent virtual resource management functions are in place to monitor, provision, federate and fence all the resources in the network, I have to agree with the erudite blogger StorageBod that it would be a better way of doing things.  The questions are can we get to that point and how long will it take?

Anyway, about a year ago I wrote a blog post at Dell stating my LAN/SAN separatist opinions and stated that I thought FCoE was stupid and that native iSCSI over Ethernet would be less expensive and complex.  There was a push back about this from within Dell because the company was preparing to support Cisco's Unified Fabric initiative and FCoE - in large part to support EMC's roadmap.  This episode was one of the first indications to me that maybe I wasn't such a good fit for Dell.  

Now today, I wonder how Dell feels about their decision to back Unified Fabric.  I just read a blog post from Dell (thanks to Stephen Foskett for tweeting) that says seems to indicate that it wouldn't make a difference. What else can they say or do now?  I know with certainty that if I had a vision for enterprise virtual networking a year ago, like Cisco laid out yesterday that my opinions about unified networking would have been significantly different.  Regardless. I'm happy where I am today.

It's going to be interesting to watch the relationship between Cisco, Dell and EMC in the months to come. What happens when Cisco, Dell and EMC are all working on a sale and the server business goes to Cisco? Does that mean the next time Dell would lead with EqualLogic and pure iSCSI? Then what happens after that to Dell's Unified Fabric strategy?  Relationships will be strained, but will they break? This is just one example of how risky UCS is for the industry and why it is so compelling to think about.

Comments

[Charlie]

You were right, FCoE is a dumb idea. Its based on the marketing lie that storage needs a perfect network and that TCP and Ethernet are barely better than soup cans and some string.

Calling FCoE consolidated is propaganda. Consolidated would have been iSCSI: one protocol suite, one network infrastructure, one skill set.

The world will come around on iSCSI, but it will be bottom up. Remember, SNA and Token Ring were going to rule, then ATM was the great answer. FCoE will go the same way.

[marc farley]

Maybe, the problem as I see it though is that the leading iSCSI-based technology companies that I thought would push iSCSI to higher levels were acquired. Their once furious ambitions have been blunted by large companies that now want to fit them into market niches.

[Bill]

Personally I don't believe in trying to argue iSCSI vs. FCoE. They are for different use cases.

That is, of course, true if you don't believe in the one-size-fits-all philosophy in your personal and work life.

iSCSI has benefits over FC and FCoE. FC and FCoE have benefits over iSCSI. Some applications and customers will prefer choice, and tailor their storage decisions around their needs.

Those who think that we should only have iSCSI don't get it, and think that Cisco and others are out to wipe out iSCSI. It's sort of like S.F. Giants fans who have this bitter rivalry with the L.A. Dodgers, who are mostly blissfully unaware of the hatred.

The move to 10G iSCSI is non-trivial. For the most part, the silicon vendors have left the iSCSI HBA scene, giving customers only the choice of software initiators. FC (and FCoE) don't require this hardware acceleration to perform at low CPU utilization by the host and array.

For those that say "software initiator is good enough", I say "Let ME decide what is good enough, give ME choice". The market has done that, and I hope we continue to have multiple choices in the market. Choice is a good thing.

[marc farley]

I admit to having some implanted religion chips, but not conspiracy theories about technology wipe outs. The old religion predicted that Ethernet and IP would always triumph and that low cost disruptive technologies would too. But as you point out, customers want to decide for themselves. As much as I want to be correct, I am no model of perfection.