Quality of Service

[pmorr]

I am new to managing an IP video conferencing network. Can someone point in the right direction for setting up Quality of Service?

[televideoguy]

This is a topic that has been batted around the network I support for a long time. Some people say that by the time you need QoS, you really should just increase available bandwidth. While I understand what they are saying, my opinion is that QoS is preferable in networks that have video traffic, because the transient traffic that isn't managed by some set of rules is what will one day kill your CEO's video conference even if you have what looks like ample nominal bandwidth most of the time... And then the day will just gets worse from that point on...I think of it kind of like the difference between RMS and peak power, although that analogy is not perfect. But I digress.

Sometimes the video conferencing vendors help feed the confusion about QoS. For example, in Polycom's PVX desktop application config screens, there is a checkbox and text that could lead some users to believe that just by putting a check in a box they are enabling QoS on their networks, which just isn't true.

Another thing to consider is that QoS implementation people sometimes talk about network-based QoS and application-based QoS. The three main types of network QoS I'm aware of are RSVP, DiffServe, and IP Precendence. I believe that DiffServe is currently emerging as the preferred choice for many QoS implementations, but I may not be correct on that. In order for network QoS to really work, it needs to be applied as an end-to-end or sytem-wide solution. So all the video conferencing and network devices need to be using the same QoS method and need to be configured appropriately. This can be difficult or impossible to manage if you communicate across multiple networks or with a lot of one-off endpoints that are off-net.

Examples of application-based QoS would be Polycom's dynamic bandwidth allocation, video error concealment, and lost packet recovery mechanisms. Tandberg and other vendors have similar features that can compensate for some types of errors either in a single-ended mode or in conjunction with a similarly configured device on the remote end.

The other point I should mention is that a lot of our member institutions are now using Packeteer or NetEnforcer devices to get a better idea of what applications are in use on their local networks and to strictly control and define available bandwidths by application and/or protocol. I understand a lot of routers (and possibly layer 2.5 switches as well) are also beginning to have some of these capabilities. This approach has worked well for a lot of our users in managing their local networks, but they have to devote significant time to keeping their bandwidth management tools up-to-date and properly configured.

If you go to Google, Copernic, or SpaceTime and type in 'H.323 QoS' you can find a wealth of information about QoS. One thing I would suggest is to work closely with your network admin on any plans to implement network QoS, as (s)he really needs to understand the scope of what you are trying to do and the implications for supporting all of the applications and devices across the network. I'd be interested in hearing other's thoughts on this topic as well.