Cisco Blogs

Unified Fabric: Getting From Here to There

March 31, 2009 - 6 Comments

When discussing unified fabric with customers these days, the conversation is increasingly shifting from “why?” to “how?” Here are some of the slides I will use to brief customers on the “how”. The important thing to note is that the steps I highlight are simply waypoints on the journey to a unified fabric. The rate at which a customer moves is really dictated by their specific circumstances–I see customers chasing this goal aggressively and I also see others moving at a more sedate pace. For a little more detail, you can also pull down the related solution overview.

In an effort to keep conversations fresh, Cisco Blogs closes comments after 60 days. Please visit the Cisco Blogs hub page for the latest content.


  1. John:Agreed that forward progress in the data center is not always a linear process. :) I always relate the story of my CCIE lab (1994) when I had to contend with IP, IPX, LAT, AppleTalk, Ethernet and Token Ring, but today, you walk into most data centers and you will see IP and Ethernet --the change did not happen overnight, but that is eventually where the market settled for a number of reasons.As far as the 40Gb IB, I could counter that 40GbE is on the radar, although I think most customers have their hands full navigating from GbE to 10GbE. I think both technologies will sell into their existing constituencies, although the case could be made that Ethernet economics will snag more IB customers than going the other way.If we step back a bit, I think successful virtualization and cloud computing strategies will depend on more than just fast pipes"".Is this the right direction--well, we think so, but only time will tell. :) Either way, I am sure it will make for some good blogging.Omar"

  2. Aneel:Well, we announced EoL/EoS support for the SFS family a few months ago, so, hopefully, there is no secret where we are investing our efforts. :)We see a home for IB in ultra-high performance environments--think esearch center"" not ""data center"". But even in the traditional home for IB, the high performance compute cluster, more than half of the Top500 Supercomputing sites use Ethernet (56% vs 28% for IB).Within the data center, we see Ethernet as the preferred transport because of its economics, continued evolution and broad support. To be clear, Cisco's unified fabric vision is built upon 10GbE and its decedents. While it is possible to build a unified fabric with IB, I would really see this as a bridge solution for an existing IB shop and not an area where we are focused.Omar"

  3. John:Our vision for a unified fabric is based on helping customers reduce their TCO through simplified infrastructure and in improving the functionality and agility of their data center network. To some degree, this is protocol agnostic, since customers could accomplish this through iSCSI or through FCoE, depending on their circumstances. In fact, I believe the typical data center will end up supporting all three (iSCSI, FCoE and FC) for some time.Given our typical customer already has a considerable investment in Ethernet and Fibre Channel, along with the management and operational structure to support it, that is where we are focusing our efforts consolidation and convergence efforts.Regards,Omar

  4. Omar:It is a rare situation in IT when data centers truely converge on one thing and progress that way from that moment forward. Mellanox just announced 40Gb IB. The need for performance will be driven by virtualization and cloud compute architectures. Hope Cisco isn't making a strategic mistake here.John

  5. Omar.. you didn't answer the IB question clearly and I don't think you want there to be a tacit assumption that your answer is that Cisco’s vision of the Data Center Fabric appears to leave no room for InfiniBand."" Or do you?"

  6. Omar:Cisco's visison of the Data Center Fabric appears to leave no room for InfiniBand. True?John WebsterIlluminata, Inc.