Quality of Service and Resource Allocation are critical.
IMS requires the carrier to have greater control over their network resources:
Offering multimedia services on fixed-line networks and mirroring the fixed-line experience through mobile networks requires several capabilities not yet present in either network. Chief among these is the regulation of QoS (Quality of Service) to maintain sufficient throughput within minimum delay limits appropriate to each session.
In the context of packet networks, Quality of Service (QoS) refers to the ability of the connection to meet certain specified service commitments. Different types of traffic and different services require different QoS levels due to their sensitivity to various transmission errors. The major impairments experienced on packet networks include:
- dropped packets – routers may drop some packets when buffers are full
- out-of-sequence packets – packets arrive out of order and must be re-sequenced
- latency (delay) – packet transit time is delayed due to excessive retransmissions
- jitter – signal variations that can introduce timing errors or bit errors
- bit errors – bit corruption that can accumulate and degrade message integrity
The quality of certain services is much more susceptible to some types of errors than others. The real-time nature of voice communications over packet, including VoIP, makes the speech quality very susceptible to latency (delay) more so than to bit errors. Dropped packets may not affect some services, such as voice, but would ruin the integrity of a file transfer. Therefore, the network must identify the different types of traffic on the network and ensure that adequate resources are available to meet the different needs of the different types of traffic.
To control QoS, networks must establish a set of traffic classes for services based on the type of impairment that is most important to control. These traffic classes can then be prioritized across the network to ensure the required resources are devoted to the service to preserve the quality needed. In general, QoS traffic classes prioritize the following two types of services:
- Time Sensitive Services – Services such as Voice or Streaming Media which are very delay sensitive but are able to tolerate some packet loss and not lose their value.
- Content Sensitive Services – Services such as file transfers and interactive services which are not as time sensitive but are very sensitive to bit errors and packet loss.
UMTS specifications define four traffic classes for the management of QoS:
- Conversational – VoIP, voice
- Streaming – Video and Multimedia Content
- Interactive – Web browsing
- Background – Email and file transfers
The major distinguishing factor among these four classes is whether they are time-sensitive or content-sensitive. Conversational and Streaming classes are real-time transfers of data and are most time sensitive while Interactive and Background must preserve the accuracy of the bearer channel payload content.
In the Internet, five traffic classes have been defined to control QoS characteristics of traffic compared with the four classes defined for UMTS:
- Premium Constant Bit Rate (PCBR) – VoIP, voice
- Premium Variable Bit Rate (PVBR) – Video and teleconferencing
- Premium Multimedia (PMM) – Audio/Video downloads, adaptive video
- Premium Mission Critical (PMC) – Transactions, database queries
- Best Effort (BE) – Email, all else.
It is very important to note that it is not enough to allocate resources to ensure that the bearer (data) channel can achieve the prescribed Quality of Service requirements. The control signals passing among the nodes in the Transport Plane and the Control Plane must also be QoS managed to ensure that the entire session quality is achieved.
There are two primary means for providing enough network resources to meet all the QoS needs present in the demand on the network. One way is through gross over-provisioning where large amounts of resources (queue size, bandwidth, diverse routes) are provided to ensure that the peak traffic experiences maximum quality. Obviously very uneconomic in large scale networks and requires forecasting of peak demand.
The other approach is to control quality and allocate resources through a reservation process by which the initiation of each session includes a request for the required resources needed based on the application of the session. Resource requests can then be accepted or rejected based on available resources at the time. Premium services requiring premium resources can be charged at a premium as well. The concept of Differentiated Service (DiffServ) for the internet is favored as the means to regulate resources on a QoS basis. DiffServ routers would prioritize traffic based on QoS requirements and the available bandwidth of the network. At times of congestion, lower priority services would be dropped in favor of higher priority or premium services. An abundance of dropped or denied services due to inadequate resources is a clear sign that capacity growth is required in the network.
For more information see: