Quality Of Service

      The term quality of service (QoS) originated in the field of communications and was used to describe the technical characteristics of data transmission. The notion can be generalised in order to describe any service that a system offers, as done for example in [VKBG95]:

''Quality of service represents the set of those quantitative and qualitative characteristics of a distributed multimedia system necessary to achieve the required functionality of an application.''

The following criteria, adapted from [Ste95], are used to classify the quality of a service offered by an active resource:

Throughput.

Throughput is defined as the product of the size or number of the data units and of the rate at which the units are processed.

Delay.

The local delay is the time a resource needs to complete a task; the global delay is the total delay a data unit needs for travelling from its source to its sink, thereby making use of several resources.

Jitter.

Jitter is the maximum allowed variance in inter-arrival the time interval of the data.

Reliability

Reliability issues comprise error frequency and possible error types as well as error-detection and error-correction mechanisms.

Using the notation of Husemann [Hus96, p. 16--21,], a QoS requirement q necessary for processing a single medium is defined as the n-tuple of values , which parametrise the required usage of resources :

The set of QoS requirements of all k substreams constituting a multimedia data stream defines the  quality of service profile Q of the data stream:

Similarly, the QoS profile Q of an application a comprises all QoS requirements that are needed to handle the involved m media:

To assure timely presentation or manipulation of multimedia data, an application requires either guaranteed services, for which strictly observed  quality-of-service guarantees are given by the service-providing system entity, or fairly reliable  predictive services, which behave as specified for the most part. In both cases, an application must consult the  resource manager that is responsible for a particular system service to make a  resource reservation.
An application reserves resources by performing the following steps:

1. Estimating the user's subjective QoS wishes.
2. Mapping these estimations to a technical QoS profile.
3. Negotiating a contract about the QoS values with the resource management.

A resource manager handles a QoS request in four phases:

1. Schedulability Test. The resource manager checks the (worst-case or statistical) resource utilisation and decides whether the additional request can still be met.
2. Quality-of-Service Calculation. The resource manager estimates the best possible performance of the requested service that can be guaranteed or sustained for most of the time.
3. Resource Reservation. The resource managers and all other entities involved store the new allocation with the existing ones.
4. Resource Scheduling. During run time, incoming events and commands are scheduled according to the contracted guarantees.

The actual QoS values in a system can vary over time. A system must therefore permanently monitor its state and take measures to meet the guarantees, or must notify the clients about insufficiencies. The system components should also be prepared to renegotiate the contract if a user decides to alter the requirements.