2

Multi-Node VSS Core In this diagram we have 4 Cisco 6500 switch, which are connected via standard 10 GB VSL links. The questions are:

  1. Since we could only have one Active and one Standby, which one of these switches will be VSS Active and which one will be VSS Standby?

  2. Will there be just one Active VSS switch for the all of the 4 switches, or Active/Standby is a per link concept? For example, between Switch 1 and 2 one will become Active and the other one will become Standby. Will the same thing take place between the other pair of switches?

  3. If question number 2 is a correct assumption, let's say that between Switch 3 and Switch 4, Switch 4 becomes Active and Switch 3 becomes Standby. At the same time, if we choose to have Switch 2 as Active between Switch 4 and Switch 2, what will happen then?

  4. How many Active switches is it possible to have here?

  5. What is the common practice, or your personal experience, with regards to utilizing all these switches in a design like this?

I'm so sorry, my mind is so puzzled that I asked all these questions in a row. Actually, it's my new task in hand and I have never ever done such thing before. Please share any kinda link, web page, or reference which you think may be helpful.

Brett Lykins
  • 8,278
  • 5
  • 36
  • 66
Mina
  • 21
  • 1
  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you could provide and accept your own answer. – Ron Maupin Aug 11 '17 at 16:23

2 Answers2

3

VSS only allows you to aggregate two physical chassis at a time. Remember there are really only two main drivers to even using VSS (and many arguments against it depending on who you ask):

  1. Etherchannel aggregation across multiple physical devices for hardware redundancy.
  2. Ease of management/configuration by consolidating the two chassis into a single point of management.

In a "box" design like this, comprising four switches, you would use VSS on each pair separately. Then use the normal network protocols (Spanning Tree, routing, etc) to manage traffic to/through the other pair.

Based on your questions, I thought it might be good to clarify a few basic components of VSS.

First, While it does logically aggregate two separate physical chassis, the way it does so for the Data Plane and the Control Plane are different. In a VSS pair the Data Plane is active on both switches, however from the Control Plane/NMS perspective only one switch is active and responding. This is what gives VSS it's single point of management.

Secondly, the VSL interfaces are only used among a single pair of switches. To cable two pairs of VSS Switches together (for a design like you reference), you would simply use existing ethernet ports and configure 802.1Q trunks or routed interfaces as needed in your design. Each of the two VSS pairs would appear as independent switches to the other. No multi-chassis Etherchannel or single point of management functionality is possible between the two VSS pairs, they are for all intents and purposes two separate switches.

See this very helpful VSS Q/A document that Cisco has put together, or the Cisco Solution Overview document for VSS for more information.

Community
  • 1
Brett Lykins
  • 8,278
  • 5
  • 36
  • 66
  • So basically here in this diagram I could have just VSS between each of 2 switches, for example between Sw1 and Sw2, and also between Sw3 and Sw4, and the links between Sw1 and Sw3 and also link between Sw4 and Sw2 can not be VSS, am i right?is doing so a common practice? do u have any recommendation/ suggestion ? – Mina May 03 '15 at 05:34
  • @Mina that is correct, VSS will just exist between SW1/2 and SW3/4. As for recommendations/suggestions and if this is common practice, those are very very broad questions. I'd recommend getting much more familiar with VSS and asking any specific questions that come up as you learn. – Brett Lykins May 06 '15 at 01:38
-5

That design and equipment choice is sub-optimal. Cisco Nexus and/or 4500x switches are better value for 10G aggregation. Access cross-chassis etherchannel to core is what you should do.

Ronnie Royston
  • 4,399
  • 1
  • 12
  • 28
  • 1
    unfortunately those are what we got, and we must make it work somehow – Mina May 05 '15 at 08:27
  • 1
    Please consider editing this answer as it has a number of problems. First, it doesn't address any of the questions asked by the OP. Second, while statements are made about what should be done, there is nothing provided to establish this is more than opinion nor does it give any basis for the reason why the statements would apply to the situation. – YLearn May 06 '15 at 03:18