Traffic Shapping on ESW500 switches

Question

Hello all,
I am faily new on the traffic shapping / bandwith control department, and have been asked by a client to be able to do some traffic shapping in their main office.
I've read some documents on the ESW500 switches and wanted to make sure that my understanding of the technology and product is right.
My client wants to prioritize bandwith per port depending on the bandwith usage needed. Some users needs a lot of bandwith for critical applications while others
uses a lot of bandwith for not so critical applications...
From what I've read, I should be able to replace his switch with an ESW500 switch and do some traffic shaping. He has a pretty basic setup with about 30 computers connected to a switch which is connected to a router. As mentionned, if I replace the switch with an ESW500 switch, will I be able to limit per port bandwith usage both inbound and outbound?
Is there also some kind of monitoring tool I could use to see how much bandwith used per port (not required but would be a nice to have feature).
Thanks!
Matt

Answer 1

Hello Matt, I hope this finds you doing well.
The amount of bandwidth you need is directly related to the performance you require.  What performance do you require? 
Here is one example (for voice)
150 msec one way delay
Less than 20 msec jitter
Less than 1% packet loss 
With requirements for this, you will have to have the right amount of bandwidth per call and queueing control, or else the call not sound right or may even be dropped.
Another example might be for Microsoft's RDP:
Some web posts mention that 30kps is needed, and some say 100kbps is acceptable.
So this is a bandwidth target, lets say 100kbps per user for this example.  If you have 10 concurrent sessions, then this is a max amount of bandwidth needed around 1Mbps.  With statistical multiplexing, you should not need this much, since not all clients will burst or use the max amount of bandwidth at the same time.  You can probably get by with less ...
The point I am making with this, is once you know how much bandwidth you need, then you can begin to engineer your network so that critical applications are able to function and user productivity is kept high. User productivity is so very key, and it is very good to see you looking into this as you are Matt.  Good stuff Matt.
Do the application vendors have suggestions for you for how much bandwidth their apps require?
If the application vendors do not know how much is needed, then using Cisco devices, you can define the performance you require and the Cisco device will respond with the required bandwith per queue.  This is called 'corvil bandwidth' or bandwidth estimation. Using this bandwidth amount, you can configure your QoS settings on all your devices.
Here is a link and I would suggest to poke around some:
http://www.cisco.com/en/US/docs/ios/qos/configuration/guide/qos_bndwidth_est.html
Application performance is an end to end perspective and consideration.  You need intelligence at the edge, core, and remote edge.
You can also look at the interface statistics to understand what an end user is using when accessing an application.  This can be an inaccurate method, but it might help you in your understanding.
With all of this said, most folks do not notice a lot of congestion on the LAN or from one LAN port to another LAN port.  Usually problems occur at bottlenecks and uplinks.  Where are the bottlenecks in your network?
Once you know where these are, then you can apply a QoS policy to protect the important traffic and ensure application performance meets your needs.  What are your needs?  See the above for some thoughts.
Does this make sense?
The ESW is a great switch, and QoS can be easily configured to meet your needs.  I think a harder question is ... what are your needs?
Here is a link for the ESW switch.  Click on resources to find the docs provided with this switch:
http://www.cisco.com/cisco/web/solutions/small_business/products/routers_switches/esw_500/index.html
HTH,
Andrew Lee Lissitz