Stumble It!
If you do some research on the subject you will find a wide variety of definitions. If you ask your IP Telephony vendor about their Admission Control capabilities you may just get a blank and puzzled look!
What follows is my own definitions and explanations of why Admission Control should be part of all IP Telephony systems.
What is Admission Control?
In the context of IP Telephony, and at its most basic level, it’s the ability of the telephony system to allow or disallow a voice path based on resource availability. The availability of the resource can be based on an actual real-time calculation or based on a pre-defined policy.
For example a system could refuse to setup a call based on the fact that a real-time calculation has revealed that the destination network link for the voice path does not have sufficient bandwidth to support the call. Alternatively, a call could be disallowed simply because the maximum number of calls configured by the system administrator has been reached.
Why is Admission Control Necessary?
There are many reasons why admission control is required and some are listed later in this article. The most important reason however, is to preserve voice quality. In a business class voice system, allowing a call to be placed over a network link that can not guarantee sufficient bandwidth is not acceptable. Some might argue that all you have to do is to design your network so that sufficient bandwidth is always available. In most instances this not accurate or practical because:
* All network links have a limited amount of bandwidth and are subject to congestion. Many people are still under the illusion that if bandwidth capacity is much greater than average data traffic requirements that congestion will not occur. This is simply not the case. In most practical applications traffic can and will use 100% of available bandwidth even though it may be for very brief periods.
* Prioritizing with QoS mechanisms is still not a guarantee that voice traffic will not be subjected to congestion. This is true even if the voice has been given absolute priority over all other types of data. Why? Because no amount of traffic prioritization can prevent over subscription of bandwidth. If the total amount of voice traffic exceeds the available bandwidth, packets will eventually be dropped. When that happens voice quality will be degraded for ALL calls over that link not just the last one that caused the amount of available QoS enabled bandwidth to be exceeded.
* There are many instances where there is a requirement to ensure that a minimum amount of bandwidth always be available for data traffic. For example, a company may decide that 1 Mb/s of bandwidth always be available for normal data traffic across a T1 link between two locations. Not counting potential routing or other overhead, that would leave approximately 500 Kb/s for voice traffic. Without admission control it may not be possible to ensure that voice never demands more than the available (by policy) bandwidth. Sure, the link can be engineered not to drop any voice traffic but if it exceeds the 500 Kb/s, it will do so at the expense of other traffic.
* Some will contend that the solution is simple: design your system so that maximum amount of potential voice traffic can never exceed the amount of QoS enabled bandwidth. While this is possible, I don’t believe that it is practical for most implementations that involve limited bandwidth WAN connections. Unless there is a very large difference between the maximum amount of simultaneous voice traffic vs. the amount of QoS enabled bandwidth as might be encountered in a pure LAN environment, the pre-engineered approach will be difficult to sustain over time. Besides, as I will describe later, IP Telephony deployments may introduce unexpected voice bandwidth requirements.
* Still others would say that it is better to allow all calls even if there’s a chance of voice quality degradation than to block a call attempt. Explain that one to the president of your company who throws the phone against the wall because he couldn’t hear an important conversation.Types of Admission Control
Earlier, I gave a very basic definition of admission control. Unfortunately, this basic functionality is all that will be available from most vendors if they provide it at all. I believe there’s a need for more sophisticated admission control functionality as described below.
* The IP Telephony system should understand how much voice traffic is being consumed over all network links. This is not simply a matter of adding the total number of calls and multiplying by a fixed number. Rather it should be aware of the actual bandwidth consumption based on whatever mix of Codecs are in use and traffic overhead such as encryption. The system should then compare the actual bandwidth consumption to a preconfigured threshold to determine if it should allow a connection to be established. A couple of vendors actually support most of this level of functionality.
* Imagine a remote office at the end of a limited WAN link. All call signaling / control is located at the main office location. Also, the voicemail system is located at the main office with no local media streaming capabilities (the subject of a future discussion). The previous evening, the president of the company has left a lengthy broadcast voicemail to all employees. As the employees at the remote office arrive for work they notice the message waiting indicator and they start listening to the voicemail. Most systems are designed to accommodate a maximum number of expected calls based on industry standard baselines. So while standard engineering may have resulted in a design that has enough voice bandwidth to accommodate the fact that say 20% of the employees will be on the phone at the same time, how will it cope if 80% of the people are listening to the voicemail simultaneously? While some vendors recognize and impose limits on regular telephone calls, I’m not aware of any that can recognize and limit this scenario. At least, not the last time I looked. The system should have the ability to limit the maximum number of simultaneous voicemail streams to the remote office. Not only to ensure voice quality but to make sure that a minimum number of regular calls can occur during the event.
* Let's say that you’re a service provide with no ability to limit the maximum number of simultaneous PSTN calls by a single customer. I subscribe to your service and you’ve engineered your PSTN resources based on a 6:1 ratio. There’s lots of bandwidth available between my location and your hosting centre. I open a contact centre and my use of your PSTN resource is more like 1:1. I am now using an unfair share of your resource and not paying for it. The system should allow you to limit not only the number of calls but the type. All IP calls between my two branch offices for example, would not impact your PSTN resource and you may not wish to limit those.
Just which component of an IP
Telephony system should be responsible for admission control is
arguable. Some will insist that it is not even necessary in a properly
engineered deployment. My opinion is based on real life experiences and
I’m convinced that admission control is just as important as QoS
mechanisms in enterprise telephony.
Oh and by the way, when
admission control mechanisms refuse a call setup, it would be nice if
the system would provide a voice system announcement that explains why
the call attempt failed rather than a nonsensical fast-busy tone!
Rick McCharles
Telcommunications Consultant, Toronto, Ontario Canada
http://www.ric.ca/
