The QoS Challenge

The challenge of QoS is not introduced by wireless networks alone, but it was realized with the introduction of new high-bandwidth applications on the Internet. Normal IP data services, referred to as background or best-effort services, like email, Web browsing, FTP, and telnet sessions can work fine without a need for QoS. As new applications like voice over IP, multimedia streaming, and other bandwidth-hungry applications come into existence, the need to manage, control, differentiate, and guarantee the desired service levels for the duration of the communications has become an important issue. The user perception of quality is determined by end-to-end factors like latency, jitter, throughput, bit-error rate, and bandwidth. QoS management and the associated traffic engineering mechanisms together provide desired service levels.
Providing end-to-end QoS between two communicating endpoints is not a trivial issue as they are separated by networks owned and operated by several operators. There must be common understanding of QoS service levels between the users and the network and across the network borders. One means to achieve this is by establishing service-level agreements (SLAs) between users and the network in the form of subscription levels and between network operators to enforce service guarantees. Presently, on the Internet, SLAs are established for the aggregated traffic from all users across various ISPs and backbone service providers to provide a guaranteed level of service usually in the form of uptime, bandwidth guarantees, and delays.

However, QoS differentiation occurs on an application or service basis or even on per user basis, meaning that all services are treated equally and users cannot request for a higher QoS for a VoIP call to enjoy a better communication experience. There is also an effect due to mobility on QoS. Frequent changes of CoA due to mobility and handover make it difficult to maintain the same QoS levels from one point of attachment to another. Over the past years a lot of work has been devoted to understanding, defining, and developing QoS architectures and protocols. The IETF Integrated Services (Inst-Serv) group has developed an integrated service model and QoS framework for QoS provisioning on the Internet.

IETF has also specified another method for QoS provisioning, called the Diffserv. It offers a scalable solution by not requiring establishment of per-flow states in the network, but by aggregating the flows into predefined service levels. Packet classification and per hop behavior (PHB) are the building blocks of the Diffserv networks. Broadly, two different PHB are defined expedited forwarding (EF), for delay-sensitive real-time data, and assured forwarding (AF) for noncritical data. There are further four divisions within the AF class. Further, DiffServ does not require prior end-to-end signaling, but the source node can perform packet classification by appropriately marking the IP packets with desired DSCP codes, corresponding to desired QoS levels.