On Saturday 01 November 2008 15:35:08 Dennis M S wrote:

> Currently focus is on the campus LAN..

How have you deployed your campus LAN? Is it a Layer 2 
Ethernet switched or Layer 3 IP routed network?

Whichever type you have would help you decide how best to 
use your equipment resources effectively, with regard to 
QoS.

> as proof of 
> concept(actually my faculty vlan's~6,deployment will be
> in the public vlan...the intranet ip video is an add on
> feature not a priority as for now)...when people do have
> a reasonably pleasant time using the set up..this will
> encourage wide spread usage.

The thing to remember is that QoS generally takes effect 
during times of congestion. When output interfaces are not 
congested, do not experience sudden spikes or do not 
surprisingly run out of packet/frame buffers, QoS features 
are typically not in effect (although they are monitoring 
the state of the interface).

This does not mean you don't need QoS. You still do need 
QoS, even on a 10Gbps network, because the sudden spike 
that exhausts all buffers can cause your sensitive 
voice/video traffic to have a potentially high drop 
probability.

I'm not sure what vendor equipment you have in your network, 
but whatever the case, if you have the budget, perform QoS 
features in hardware (queues). But if not, and your traffic 
levels are not that high, doing this in software (queues) 
is fine as well, as long as you watch that CPU bar.

Your QoS strategy should be based on DSCP (Differentiated 
Services Code Point) - DSCP defines a PHB (per-hop 
behaviour) where each router/switch in the node provides 
the QoS services configured for a packet/frame when it 
first entered the network, up to the point when it egresses 
it:

http://en.wikipedia.org/wiki/Differentiated_services

Voice should be treated as low latency (EF or Expedited 
Forwarding as defined in DSCP). If your equipment supports 
low latency queues, you should place your voice traffic in 
there. These types of queues are serviced first before any 
other types of queues on the interface. This means that 
it's generally recommended not to assign more than 33% of 
the interface bandwidth to the low latency queue, otherwise 
you risk starving other queues and/or traffic flows.

Interactive/real-time video (along with audio) should also 
be assigned to a low latency queue, e.g., a live video 
teleconference.

However, video-on-demand (which as you say would be a 
requirement), does not have to enjoy the services of a low 
latency queue. You can provide assurances that traffic will 
be delivered as long it remains in-profile, i.e., it does 
not exceed CIR (or PIR, if configured). DSCP calls this AF 
or Assured Forwarding.

Since VoD isn't real-time, users can always go back to it in 
case the network is facing congestion. So you don't need to 
give it the same treatment as real-time video.

> The P.A(public address) system is already in
> place..recording lecture's shouldnt be an issue...,the
> bare bones functionality is to have people call the voip
> server from their gsm mobile's and listen to stuff e.g
> announcements,recorded lectures..etc,and ensure calls out
> to gsm from the voip setup .

This should be fairly straight forward then. Since each user 
is coming in via GSM, you just need to make sure your VoIP 
server can keep up with the load associated with 
transcoding - since GSM wouldn't be IP.

If, however, you are trying to provide the stream over IP, 
then the QoS considerations I mention above would come into 
play.

Cheers,

Mark.

Attachment: signature.asc
Description: This is a digitally signed message part.

_______________________________________________
LUG mailing list
[email protected]
http://kym.net/mailman/listinfo/lug
%LUG is generously hosted by INFOCOM http://www.infocom.co.ug/

The above comments and data are owned by whoever posted them (including 
attachments if any). The List's Host is not responsible for them in any way.
---------------------------------------

Reply via email to