As you probably know, Cisco's long-standing recommendation is to use Low-Latency Queuing (LLQ) for VoIP. Low-Latency Queuing is simply Class-Based Weighted Fair Queuing (CBWFQ) with the addition of a single priority queue. In most places where this is deployed, the only thing placed in the priority queue is VoIP traffic. However, what you may not realize is that you can put other types of traffic in there.
The confusion, I believe, stems from a misunderstanding of the terms "queue" and "class"; some people may believe they are the same thing, while others may realize they are different, but assume there is a 1 to 1 relationship.
The thing to understand is that in CBWFQ you can have a lot of classes. When you turn CBWFQ into LLQ by creating a Priority Queue (PQ), you can actually put up to four classes in the Priority Queue. But remember, there is still only a single Priority Queue.
It's also important to realize how these are treated. Each of your four classes has its own bandwidth allocation. They are policed and rate-limited individually. Note also that if there is no congestion, they're all free to use whatever bandwidth is available. Packets are only dropped in the event of congestion.
Now, this feature may be useful to you if you have more than one type of traffic that needs the guaranteed treatment of a strict queue, but put some thought into the total bandwidth you want to allocate to the Priority Queue before you implement it, and decide how you want to divide that bandwidth up between the priority classes.
Tom Lancaster, CCIE# 8829 CNX# 1105, is a consultant with 15 years experience in the networking industry, and co-author of several books on networking, most recently, CCSPTM: Secure PIX and Secure VPN Study Guide published by Sybex.