Options

backplane question

marcusaureliusbrutusmarcusaureliusbrutus Member Posts: 73 ■■□□□□□□□□
Hi. Can anyone please explain the actual, practical use of a backplane? Am i right to say that a backplane is the forwarding bus of a router/switch?
Is the backplane capacity important to consider in typical networks, if yes, then why? Is there a reference available, some sort of guide as to what backplane capacity to use for certain lan designs(a 24 switch connected to 24 workstations, is there a minimum backplane requirement?

Thanks in advance.

Comments

  • Options
    kryollakryolla Member Posts: 785
    Im not a hardware engineer so excuse me if I dont sound too intelligent. My understanding is the backplane is how the cards in slots talk to the sup and to each other i.e in a 6509 you have 9 slots which consist of the supervisor which is the brain/controller of the chassis and you have 7 other slots for input/output devices like ethernet blade. The way these slots talk to each other and to the supervisor card is via the backplane. The backplane also has capacity limits which should be taken into account if you have the box maxed out. The bus is how the different processor within the card talk to other peripherals like memory modules, etc but within the card itself. I have to go over the IOS architecture book again. But that is the laymans version of what a backplane is.
    Studying for CCIE and drinking Home Brew
  • Options
    marcusaureliusbrutusmarcusaureliusbrutus Member Posts: 73 ■■□□□□□□□□
    Hi kryolla,

    Thanks for the prompt reply. How would i know whether the backplane capacity is enough for a certain type of network or whether i need a switch with a greater backplane?

    Thanks in advance.
  • Options
    kryollakryolla Member Posts: 785
    its in the spec but if it is for a fixed chassis like a 3550 or 2950 then I wouldnt worry about it but look at the spec for like a 3550 and it will tell you how many packets it can switch etc. Only for bigger boxes like 6500, 7600 etc would the backplane capacity be a concern. To be honest I would be more concerned about CPU util, memory util, interface buffers, etc instead of backplane constraints
    Studying for CCIE and drinking Home Brew
  • Options
    Forsaken_GAForsaken_GA Member Posts: 4,024
    As kyrolla mentioned, for your fixed modules, the backplane isn't much of an issue, you just choose the switch that can handle the throughput and pps you need and use it, you can't upgrade it.

    For modular chassis, things are a little different. I'll use the Catalyst 6500 as my reference point, since it's the one I'm most familiar with -

    The Cat6500 actually has two backplanes. The first is the shared backplane, which can handle 32Gbps. This is a shared bus that all line cards can use.

    The second backplane is a high speed switching fabric. With the use of a switching fabric module, and with linecards that can use the switching fabric, you could get speeds of up to 256Gbps. Using a Supervisor7203B or Sup7203BXL, the fabric can support up to 720Gbps (this was a theoretical limit the last time I checked, as there was not enough room in any of the chassis to actually be able to put 720Gbps through it, though it's been a few years and this might have changed).

    So in the case of a 6500, your backplane usage is pretty important. If you have a Sup720, but you're using line cards which aren't fabric enabled, it's pretty much a waste, as any communication to and from those line cards has to go over the shared bus, thereby effectively removing the advantages of having a fabric. You have to choose your modules carefully, as some connect to the shared bus and the fabric, some connect only to the fabric, and some connect only to the shared bus. And yes, the fabric enable line cards are *very* expensive. Every time I've ever handled one, I had to remind myself that these were worth more than my car, so I'd better not break the damn thing.
  • Options
    marcusaureliusbrutusmarcusaureliusbrutus Member Posts: 73 ■■□□□□□□□□
    Thanks guys for explaining. I really appreciate it. However, regarding throughput, does it mean whether the ports on my switch is fastether or gigabit ports? I have checked the performance data sheet for the 3750 and the 2960 but i don't understand what the TS, TT, TC stands for.

    Feature3560- 48-TS G-24TS G-48TS 8PC

    Forwarding Rate 6.5 Mpps 13.1Mpps 38.7 Mpps 2.7 Mpps

    Feature296060-24TT 60-24TC 60-48TT 60-48TC 60G-24TC 60G-48TC

    Forwarding Rate 6.5Mpps 6.5Mpps 10.1Mpps 10.1Mpps 35.7Mpps 39.0Mpps

    Thank you for your time and help.
  • Options
    Forsaken_GAForsaken_GA Member Posts: 4,024
    Cisco part numbers take a little getting used to, in general, they indicate the port numbers and types of the switch, as well as the IOS image.

    For example, 3560-48-TS-S (and there's usually either a -E or a -S on the end as well) would be a 3560 with 48 10/100 ports with 4 SFP's (in general, 48 port 3560's have 4 SFP's, 24 ports have 2, there are exceptions to this though) with the IP Base image.

    A 3560G-48-TS-S would be the same thing, except the ports are 10/100/1000 instead.

    A 3560G-48-PS-E would be a 3560 with 48 10/100/1000 ports with 4 SFP's that supports PoE and has the IP Services image.

    A 2960-24TT-L is a 2960 with 24 10/100 with 2 1000BaseT ports and the LAN image.

    And to make it even more fun, the designations change with family, (ie, TT wouldn't be valid in a 3560 part number, and TS wouldn't be valid for a 2960).

    You just have to be careful with what you're ordering, especially if you're ordering from a reseller. The only reason I'm so well versed in Cisco part numbers is because I did work for a reseller for a number of years. Our sales people misunderstood or just flat out got wrong the part numbers on more than a few occasions. We also had incidents where we received them under the wrong part number, and boy was there hell to pay if someone made that mistake on ingress. Best way to get familiar with the part numbers and their variations is to make yourself friendly with cisco's sales literature.
  • Options
    marcusaureliusbrutusmarcusaureliusbrutus Member Posts: 73 ■■□□□□□□□□
    Hi Forsaken_GA,

    Thank you so much for the explanation. I truly appreciate it. I will do as you advised and try to gain more familiarity with cisco part nos. I have another question though. Is there a way to determine what minimum pps is required if i have 48 workstations with 10/100 connectivity to each port that i wish to connect to a switch? How can i tell whether i need a switch that has 32Mpps or whether a switch with a 10Mpps will do?

    Thanks again in advance.
  • Options
    Forsaken_GAForsaken_GA Member Posts: 4,024
    All depends on the type of traffic you're putting through it. If you're only putting 10 megs through an interface, but you're doing it with many small packets, then the number of packets it needs to support is larger than if you were transferring 10 mbits of large files.

    To be perfectly honest, the only time I've ever seen pps make a real difference is during a DoS/DDoS attack. Then it matters very much. Especially if it's going through an old piece of crap Foundry load balancer that can't keep up.... *ahem* theoretically speaking, of course

    Edit: Actually, I take that back.. I have seen it count on one other occasion. We have one customer who does alot of streaming video, usually to cell phones. So there are *alot* of clients and simultaneous connections. We originally had them on a Cisco 2970, and it just couldn't keep up. We moved them over to a Force10 S Series, and it held up like a champ.
  • Options
    marcusaureliusbrutusmarcusaureliusbrutus Member Posts: 73 ■■□□□□□□□□
    Hi Forsaken_GA,

    Again thank you so much for the explanation. You have completely answered my questions. I hope you all the best!


    Regards
Sign In or Register to comment.