I have been studying QoS for some time now and am familiar with the tools and commands that Cisco provides for configuring QoS.
Recently, my company has given me the task of creating a QoS policy. We will be deploying VoIP and video very soon and my task is to prepare the network for this endeavor.
While there are lots of documents describing how to configure QoS, I only find a few documents on how to design a policy. In particular, let's say that I have a packeteer or other similar device that can classify traffic and tell me what the average / peak traffic rates are. What is the proper way of setting up queuing if you have this information?
I know that for voice, I need to look at the codecs that will be used (G729 in my case) and the Call Admission Controls to determine what my voice BW requirements are (as well as the other basics such as future traffic growth, etc).
What I don't know is how to size my other traffic queues? Should I classify the traffic and just add up the average BW usage for the class? Or should I provide a BW to provide for the peak rates (I doubt I have that much BW to divide up).
I guess - what are your recommendations? Currently I am looking at LLQ to provide for my voice / video traffic (perhaps 2 priority queues, 1 for each type of traffic). Any and all help is appreciated!
Jim, first idea: go to this URL http://www.cisco.com/en/US/netsol/ns656/networking_solutions_program_home.html and read the "Enterprise QoS Solution Reference Network Design Guide Version 3.3" This guide has an excellent explanation of many of the things you've been studying, as well as the differences between devices (hardware queues, capabilities, etc.) It discusses Auto-QoS which is ONE TOOL for QoS but not a global panacea.
Second, in my real world experience, I do (almost) everything based on BW% not BPS. My basic policy defines 35% BW for LLQ for VoIP, 50% BW for production applications (Citrix, specific http/s traffic, SAP), 5% reserved for ssh/telnet/snmp/syslog/ftp/etc. and everything else in the default. Many WAN providers won't go >35% for the priority queue anyway.
I have some QoS templates for routers and switches, drop me a line if you'd like a copy.
As another good example of a policy check out the following paper:
Even if you don't have exactly the same requirements there are still many good ideas here.