QoS is (as stated) a fairly complex setup, even for some pretty seasoned Network guys. A quick writeup or PDF doesn't really do it justice. BUT, here's some basics.
A lot of times the correct method depends on the network hardware used (Cisco, HP, Extreme, Juniper etc). The Avaya systems and phones are pretty flexible in regards to how they'll interact with a network to allow it to prioritize voice traffic. It's possible you give a network admin something like "Diffserv 46" and he'll look at you like you have 3 heads.
If you're not routing to another location, then you may not be able to use DiffServ settings for a single LAN, depending on certain conditions. 802.1p is typically a Layer 2 priorization method, and you have to ensure that your phones 46xxsettings.txt file tells the phones to tag the packet with the proper value in L2QAUD and L2QSIG (usually 5 or 6, default on the Avaya is 6 but default on most Cisco stuff is 5 - pay attention!)
So regardless of how you're marking your voice packets (802.1p, DSCP) then you have to be able to coordinate that with the network admin. You'll also want to make sure that the network admin understands how QoS/priortization works on a per-port basis. Best results for voice usually entail strict-priority queues, meaning that any time an interface builds up a queue (is congested with packets), voice packets always go right through without any stopping, and data traffic has to wait in the queue until it's clear. Any other types of queues can possibly hold up voice packets, such as round-robin or weighted queues (like 2 voice packets for each data packet, something to that effect) which we do not want.
Hope this gets the ball rolling for you.
