Packet Pushers – Show 16 – Big Hot and Heavy Switches – Part 1

A detailed look at the Big, Hot and Heavy Ethernet Switches with a large crew to talk about their practical experiences on design, selection and performance of Cisco Nexus switches. The result ? We don’t think the Nexus switches are very exciting, HP Flex10 modules are a problem, and so much more.

Due to people commitments we recorded a double length show which will be released in two parts. This is Part 1 and Part 2 will be released next weekend.

VSS & vPC – the real experience

We started with a look at the differences between Cisco’s Virtual Switching System (VSS) and their Virtual Port Channel (vPC) implementation of Multi-Chassis Etherchannel and what our preferences are for choosing and implementing. We like vPC both from a design and implementation perspectives but our experiences of VSS have been very poor and we don’t choose to use it in the Data Centre. In practical way we don’t like the way that VSS is conceptually the same as a stackable switch.

We also don’t have much time on the Catalyst 4500 because they aren’t used in our designs. Also, there are too many modules and Supervisors, and the technology is confusing. The lack of architecture information is poor.

Cisco Fabric Extender switches

We had a short look at the QoS implementation on the Cat6500, and expressed our dissatisfaction with the configuration and consistency of the implementation and some surprising tips on how to solve the QoS problem.

Service Modules

We discuss the use of Service Modules in the C6500 chassis and our experiences.

The Guests

You can find Jeremy Filliben at http://jeremyfilliben.com and @jfilliben.

You can find Steve Rossen on @steve

You can find Ivan Pepelnjak at http://ioshints.info and on @ioshints.

IOS Hints Live – San Jose September 2010

You can book to join the event at ioshintsdatacenter.eventbrite.com/. There are only a limited number of seats at this unique event where Ivan Pepelnjak and Greg Ferro will both be available to discuss, review and develop your designs.

Feedback

Follow the Packet Pushers on Twitter (@packetpushersGreg @etherealmindDan@rovingengineerEthan @ecbanks) and send your queries and comments about the show to [email protected].  We want to hear from you!

Subscribe in iTunes and RSS

You can subscribe to Packet Pushers in iTunes by clicking on the logo here.

Media Player and MP3 Download

You can subscribe to the RSS feed

  • BenG

    Your knock on VSS is a complaint that it is conceptually the same as a stackable switch? That’s pretty lame in the “we don’t like it because…” department. From an operational perspective, there are very strong similarities between the two. The VPC peer link is similar to the VSS Virtual Switch Link. If you choose to go the BFD detection route with VSS, then you have the similarity with the VPN peer keepalive link. The only real feature of VSS that makes it look like a “stackable switch” is the config sync that happens between the pair of 6500s. Besides that feature, VSS and VPC have much more in common that VSS and Stackwise. As for the config sync feature, most of the folks I’ve talked to and worked with find the config synchronization that happens with VSS to be a positive over having to maintain two distinct configs on the Nexus platform.

    • http://etherealmind.com Greg Ferro

      Not quite. It was clear that most of us have experienced some very poor outcomes with VSS. We also know that expecting two autonomous operating systems to bond into a single instance (as VSS does) is inherently an unreliable system design in computing. Because VSS switching planes are attempting to act as a single instance, but the control plane is active / standby, makes it even more complex.

      Cisco track record with stackable switches, which perform the same software feat, is very poor. Most (not all, but most) people, regard Cisco stackable products as unreliable and prefer to use C4500 for high density single instance switches.

      vPC has nothing in common with VSS and is a much more proven design, not just for networking but for IT in general.

      The burden of configuring two configuration is a pain, I’ll certainly agree, but against the proven lack of reliability of VSS, it’s not a bad price to pay.

      • BenG

        Greg –

        I would argue that any time you have dual anything in a networking environment, you have two autonomous systems acting as one. I see little difference between VSS and a single chassis with dual supervisors. Whenever there is a failure, one system must come online and take over from the other. The control plane is still active/standby in a switch with dual sups…

        I’m curious how you can state that vPC is a more proven design, given the technology is still in its infancy. I’ve heard my fair share of horror stories with regards to vPC as well…

        I’ve had VSS installed and running at several customer sites with no incidence, so I find it interesting that you can claim “proven lack of reliability”. If there was proven lack of reliability, I’m sure it would be much more in evidence and everyone would be avoiding it like the plague…

        -Ben

        • http://etherealmind.com Greg Ferro

          Actually, I said that our collective experience is that VSS is unreliable. I guess it must work, Cisco is still selling it as a feature, but we don’t believe in the architecture of the feature, plus our initial experiences were negative.

          YMMV, but that seems to be our view.

          And yes, most people avoid it. At least, those people who have knowledge to comprehend how it works. It’s a dreadful hack from my POV, and hard to believe that Cisco ever released it. They spent the last ten years sand bagging everyone else who tried this type of stackable technology. I find that amusing.

        • http://etherealmind.com Greg Ferro

          Oh, and the vPC idea has been around for about ten years. Most other companies called it MultiChassis etherchannel (Nortel released it in 2002. Took Cisco a long time to catch up.

          • BenG

            The idea has been around for a while, but Cisco’s implementation flavors have not. Like most things new to Cisco, it takes a bit to work out the kinks. I think you’d find the same thing if you talk to some Nortel shops. I’ve heard a few complaining about issues they’ve had with SMLT. Its just a shame it took so long to get a MEC solution (sorry Cisco, L3 to the access layer didn’t count).

            -Ben

  • Dave N

    If there’s a download link I’m not finding it. The bit that says “mp3 download” directs me to subscribe to the RSS feed which I do, with Google Reader and there’s no download link there either.

    Also, since I’m writing, what’s with the funky, ugly, hard to read font?

    I do enjoy the blog posts and I’m curious as we use a lot a stacked 3750s and our new engineering VP is apparently a Nexus fan.

    • http://etherealmind.com Greg Ferro

      Download link is over at the http://packetpushers.net where the software takes care of all that, I just cross post the article here. If you subscrine in Google Ready you won’t be able to see the audio since Reader doesn’t display the media encapsulations.

      And the funky hard to read font is the result of your funky browser and crappy OS which is unable to render HTML5 fonts. Time to upgrade and get a better internet experience. Get Linux or a Mac and help the world be a better place.

      • Dave N

        Thank you fo r the link. I think I’m subscribing here and not there which is why I don’t see the audio link and was confused. I actually found it via Google and came back to answer my own question.

        As for the funky font I guess I’ll ha ve figure something out as my crappy OS is mandated by my employer. I guess I could read it on the Mac at home .

        Thanks for the assist.

        • http://etherealmind.com Greg Ferro

          Sorry to be anal, but I like the new fonts and feel it’s time to move on from ten or fifteen years ago.

  • http://www.gamersanon.com Andy

    I suppose it really depends on the type of network you’re doing when you talk about VSS. The only thing I would say is to consider the use cases where the network is not as wide sprawling as your discussions probably led you. I’ve been both blown away with how VSS is supporting our 24×6 facility and also notably let down.

    MEC however is awesome and actually seeing it work on 3750’s is fantastic. Trying to get it to work with 3548’s, not so much lol =p.