A few words about Nexus 7000

Filed in Nexus by on March 30, 2008 15 Comments

If you are not familiar with the innovation in Cisco Nexus 7000 it’s easy to just dismiss it as “Cisco’s answer to Force10″. The reality is, the Nexus 7000 is much more than just a “me too” 10-gig switch to compete with likes of Force10, Foundry, Extreme, and others — Nexus is a complete technology leap beyond what any other switch vendor has to offer today.

Let’s hit on some of the many points that make Nexus 7000 unique and why this switch is in a league of its own…

  • Nexus 7000 is the first switch to deliver a LOSSLESS Ethernet Fabric with Virtual Output Queueing (VOQ) and Central Arbitration that guarantee no frames will be lost entering the switch fabric. Can Force-10, Foundry, Extreme, and Juniper say the same? NO!.
    Why a lossless ethernet fabric? Data Center server I/O consolidation — The next major architectural transformation in the Data Center.
  • Investment Protection. Nexus 7000 is designed to deliver 500+ Gbps per slot and beyond with future fabric cards — no forklift upgrades. Other vendors will ask you to buy a whole new chassis and linecards to get you to that kind of bandwidth.
  • High Availability. Because Nexus was designed to deliver I/O consolidation (Fiber Channel over Ethernet), high availability was built into every aspect of the switch design from grid redundant power supplies, stateful hitless supervisor failover, In Service Software Upgrades (ISSU), to a self healing OS that recognizes failed processes and repairs itself without user intervention.
  • Virtualization. A single Nexus switch can be carved into multiple logical switches with unique and isolated control and data planes instances. Why virtualization? Infrastructure consolidation across organizational and business unit boundaries

The fact is, Nexus 7000 is not just another 10-Gig switch. Nexus is a platform that will transform the Data Center as we know it. Nexus 7000 is in a much different league than what any other switch vendor has to offer today and is a product of Cisco’s leadership in Data Center innovation and transformation.

http://www.cisco.com/go/nexus

 

Tags:

About the Author ()

Brad Hedlund (CCIE Emeritus #5530) is an Engineering Architect in the CTO office of VMware’s Networking and Security Business Unit (NSBU). Brad’s background in data center networking begins in the mid-1990s with a variety of experience in roles such as IT customer, value added reseller, and vendor, including Cisco and Dell. Brad also writes at the VMware corporate networking virtualization blog at blogs.vmware.com/networkvirtualization

Comments (15)

Trackback URL | Comments RSS Feed

  1. Will says:

    All flames and kudos aside… it sure is an ugly beast compared to a Force10 E1200!

  2. Joe Harris says:

    Will, your aesthetical taste scares me :-) Let’s let the readers decide, here is a picture of a Force10 E1200:

    http://www.force10networks.com/company/image_gallery/eseries/E1200-lg1.jpg

    and here is a picture of a Cisco Nexus 7000:

    http://www.cisco.com/en/US/prod/switches/ps9441/ps9402/product_large_photo.jpg

    The cable management section of the Force10 looks like an extremely bad hair day compared to that of the Nexus…

  3. Brad Hedlund says:

    Thanks Joe. Ugh, that Force10 really is an ugly switch. The logo on the cable management tray sorta looks like “Force0″. And it looks as if the switch is saying … “I really wish someone would acquire me, please, anybody?” :-)

  4. Darby Weaver says:

    Which came first? Looks like somebody borrowed someone else’s chassis design to me.

    When do the lawsuits start?

  5. Cavalier Poodle says:

    Twin-ax SFP+ has already been released for Nexus 5000, but why can’t
    it be used with Nexus 7000? I hope the next software release solves it, though.

  6. Brad Hedlund says:

    Max distance on SFP+ Twinax is 10 meters. It would be difficult to get a large population of servers within 10 meters of a Nexus 7000. Piece of cake with Nexus 5000 however because it is a Top of Rack switch. The majority of deployments where servers are connecting directly to Nexus 7000 will be done via fiber where the Nexus 7000 sits at the end or middle in a row of server cabinets.

    • Felix Rodriguez says:

      It really isn’t hard at all to get the 7000 within 10 meters of a large population of servers. As long as you have patch panels to each of the rows in a data center it can reach servers as far as 200 feet. The wire management on this 7000 is very organized too.

  7. One could easily do 496 servers with 10M Twinax cables, but using Myricom’s 21U 512-port switch configured with 496 SFP+ 10GbE ports. Six 42U racks on each side of the networking rack all loaded with 1U servers would be a piece of cake.

    Given that we also sell a 640-port 21U 10G Edge switch we could configure Twinax between computers and the Edge switches and fiber between Edges and multiple centralized 512-port Core switches the limit of 10M on Twinax is manageable. Actually up to 8,192 computers it’s pretty easy, after that it takes a little thought!

  8. Brad Hedlund says:

    Lets not confuse what is possible, with what is practical.

  9. Brad,
    I have been doing a great deal of work around data center consolidation and ultra low latency design. I completely agree with the FACT that the nexus should not even be spoken about in the same sentence as other switch vendors it is truly apples and oranges. This FACT is solidified when you look at technologies like ultra low latency, port cascading, enhanced low latent QOS, and then when you start incorporating ecosystem technologies like DAL and RAB Cisco puts itself in a class of its own. I do have one question however; my customers have been asking me if the Nexus switch will have the ability to power down ports when not in use. I expect to have an answer soon as we will be deploying this product in many of our accounts but for now I have no answer. Can you enlighten me on this topic since power consumption is always a hot topic?

  10. Brad Hedlund says:

    James,
    Great question! A port that is not transmitting or receiving data is using less power than a port this is operating at full line rate. So, Yes, a port that is ‘shutdown’ is using less power than a port actively forwarding data. A line card requests a power reservation from the system based on a worst case scenario, the most power it would ever need. However the power reserved by the linecard is not always the power that is actually used at any given time. This is why the Cisco Power Calculator on Cisco.com now shows “Typical Power” for Nexus 7000 configurations, in addition to “Total Power” which represents the worst case scenario. Typical Power is usually 30% less than worst case scenario Total Power due to the efficiencies saved in idle and shutdown ports. For example, the 32-port 10GE lincard reserves 750W from the system but typically only consumes 600W.

    I love the enthusiasm for Nexus!

  11. james hoffman says:

    Brad,
    Thanks so much for the quick response I am so glad I found this forum as it is difficult to get quick answers for “real” advanced technology topics. (Even when you know most of the ce and de’s at cisco) currently I am connecting a nexus solution to a wombat ticker feed for low latency I will post an update on how that goes in to coming months. Thanks again this is a great forum!!!!

  12. BG says:

    James, you seem easily amused. However, I am not so confidant that your question was answered to my delight..

    Brad, I believe it is a fare assumption that most engineers will conclude that when a port is in a ’shutdown state’ it is using less power than a port actively forwarding data.

    However I believe the question was more in line with, if ports are in an up / up state and there is 0% traffic traversing the interface will the NEXUS 7K remove the port from service and place into a Administrative / Down (Shut Down) state without human intervention? Also, in which state is power actually removed from a port?

    If yes, would this result in a less (–) power drawl on the line card as a whole, hence power reduction from the power plant?

    BTW- I think the Force10 looks like a “Predator on ludes” the NEXUS is much more sleek!
    Ref- Comment by Brad Hedlund on 12 January 2009:
    James,
    Great question! A port that is not transmitting or receiving data is using less power than a port this is operating at full line rate. So, Yes, a port that is ’shutdown’ is using less power than a port actively forwarding data.

  13. Shahid Shafi says:

    Hi Brad,

    Is VoQ applicable for ingress traffic to the switch? I am trying to understand the difference between VOQ and PFC 802.1Qbb/802.1Qaz.

    thanks,

    • Brad Hedlund says:

      Shahid,
      VoQ with Central Arbitration, such as found in the Nexus 7000/5000 and MDS, eliminates head of line blocking (HOLB) and supports lossless switching within the box. If a lossless switch gets congested, it needs to tell its link partners so they can slow down their send rate, to create a lossless network. PFC 802.1Qbb are the messages that a lossless switch sends to its link partner when its input buffers reach a congestion threshold. ETS 802.1Qaz is not a message, it just defines a switch (or network adapter) that can provide minimum bandwidth guarantees to individual classes of traffic.
      VoQ and Central Arbitration makes a switch lossless, PFC 802.1Qbb messages are how these switches communicate congestion to each other to form a lossless network.

Leave a Reply

Your email address will not be published. Required fields are marked *