Generalized QoS Processing Model
This section of the Internet Tool Survey describes the quality of service provided by the Internet. Currently the Internet offers a point-to-point delivery service, which is based on the "best effort" delivery model. In this model, data will be delivered to its destination as soon as possible, but with no commitment as to bandwidth or latency. Using protocols such as TCP, the highest guarantee the network provides is reliable data delivery. This is adequate for traditional data applications like FTP and Telnet, but inadequate for applications requiring timeliness. For example, distributed multimedia applications need to communicate in real-time and are sensitive to the quality of service they receive from the network. For these applications to perform adequately and be widely used, QoS must be quantified and managed, and the Internet must be modified to support real-time QoS and controlled end-to-end delays. The notion of QoS must be extended from the communication layer up through the intervening architectural layers to the application level.
There is no common or formal definition of QoS. However, there are a number of definitions at the communication level where the notion originated to describe technical characteristics of mainly non-time-dependent data transmission. Emerging networks such as ATM, can provide QoS guarantees on bandwidth and delay for the transfer of continuous media (CM) data.
The IEEE paper, Distributed Multimedia and Quality of Service: A Survey, provides a more general definition of QoS for applications that must communicate in real-time: "The set of those quantitative and qualitative characteristics of a distributed multimedia system, which are necessary in order to achieve the required functionality of an application." This paper also provides a model of QoS processing for multimedia systems which we will generalize below to include applications requiring Internet services. Several research groups are investigating QoS support for the WWW, in particular, researchers at BBN, the Distributed Systems Technology Centre, and Washington University. These and other QoS-based projects are summarized below.
To build QoS into a system involves
An application's QoS requirements are conveyed in terms of high-level parameters that specify what the user requires. QoS specification is different at each system layer and is used to configure QoS mechanisms at each layer. Possible system layers are
QoS specification encompasses requirements for
QoS requirements are assessed to determine if they can possibly be met. If, for example, the level of service requested cannot be provided, the user can be asked if a certain level of degradation is acceptable before proceeding further.
QoS requirements are used to derive resource requirements for entities such as computation, communication, and storage. They are successively mapped into quantitative QoS parameters relevant to various system layers that can be monitored and controlled. QoS parameters may be oriented towards
Each QoS parameter can be viewed as a typed variable with bounded values, and the values are subject to negotiation between the system layers.
To provide and sustain QoS, resource management must be QoS-driven. In allocating resources, the resource management system must not only consider resource availability and resource control policies, but also an application's QoS requirements measured in terms of the QoS parameters. To ensure the contracted QoS is sustained, it must monitor QoS parameters and reallocate resources in response to system anomalies. Prior to allocating resources, the system layers negotiate to determine if they can collectively ensure that the required QoS parameters can be consistently satisfied. Negotiation involves dynamic adaptation and the transmission and translation of QoS parameters between the layers as the layers enter into different types of agreements, e.g., guaranteed, best-effort, or predictive. If negotiation ends in agreement, the application is launched. After resources are allocated, QoS mechanisms at each layer guarantee the contracted QoS, and the resource manager guarantees the sustained availability of the allocated resources. This requires monitoring resource availability and its dynamic characteristics, e.g., measuring processing workload and network traffic, to detect deviations in the QoS parameters. When there is a change of state, i.e., degradation in the QoS, and the resource manager cannot make resource adjustments to compensate (e.g., reschedule shared resources to satisfy allocations or switch to an optimized implementation of an object/service), then the application is notified, e.g., application handlers are called. The application can either adapt to the new level of QoS or scale to a reduced level of service.
RSVP, the emerging standard for QoS negotiation over IP, is a network
control protocol for establishing and maintaining Internet integrated service
reservations that allows Internet applications to obtain both best-effort
and real-time QoS for their data flows. Hosts and routers use RSVP to deliver
QoS requests to all nodes along the path of the data stream, typically
resulting in a reservation of bandwidth for that particular data flow.
RSVP is designed for use over both IPv4 and IPv6,
the next generation Internet protocol. IPv6 offers a choice of QoS levels
beyond the single "best effort" delivery service offered by IPv4.
With these added QoS capabilities, still in the experimental stage of development,
IPv6 will provide a better range of support for real-time data traffic.
The latest RSVP functional specification can be found at CNRI ( postscript)
or USC/ISI ( text)
and includes a discussion of RSVP's design and operation. The tutorial
QoS over
the Internet: The RSVP protocol provides motivation for RSVP and details
about QoS support. See the RSVP2
project summary for details about the latest accomplishments and future
plans. For more information, see the RSVP
publication list. Collaborative research is being done at Xerox
Parc and MIT.
Research is centered around network, middleware, and OS QoS support for multimedia applications. To guarantee QoS for the transfer and processing of continuous media, ARL researchers have developed a framework consisting of four components: 1) specification of QoS requirements in terms of high-level QoS parameters; 2) the mapping of QoS specifications to resource requirements; 3) QoS enforcement by scheduling shared resources to satisfy OS allocations; and 4) a protocol implementation model that facilitates the mapping of protocol services to the mechanisms for the previous three components. See the paper A Framework for QoS Guarantees for Multimedia Applications within an Endsystem for details. More recent prototyping work in Scalable Multimedia-On-Demand via World-Wide-Web (WWW) with QOS Guarantees discusses a scalable MOD server that supports end-to-end QoS guarantees using a web-based access interface that provides complete playout control functions. In addition to real-time and end-to-end QoS guarantees, there must be QoS guarantees of the underlying services. This motivates the use of object-oriented middleware such as CORBA. Since CORBA does not support latency-sensitive applications, they have proposed the use of real-time ORBs (RT ORBs) to deliver the QoS guarantees. For further papers, see the list of author publications.
Lancaster University is one of 6 URI groups in Britain involved in research
on the management of multiservice networks; their part is to research end-system
QoS
Management. To use QoS to manage bandwidth over a multiservice network
for continuous media streams, they choose CORBA as the demonstration vehicle.
Providing QoS support involves adding QoS extensions to a CORBA-based enhanced
object model and developing various QoS mechanisms within the model.
An essential component is the QoS
mapper which maps application level QoS requirements between each component
in order to sustain the overall desired QoS. The basis for their development
is the Xerox Parc ILU
distributed programming environment.
For more information about specifying QoS requirements, see their paper
Specifying
QoS for Multimedia Communications within a Distributed Programming Environment
which examines four different approaches. A more recent paper, QoS
Support for Distributed Multimedia Communications, discusses a QoS
model to provide receiver-dependent QoS based on filtering techniques in
a multipeer communication environment. See the Ph.D. thesis Quality
of Service Filters for Multimedia Communications and the 1994 paper
Supporting
Quality of Service in Multimedia Communications via the use of Filters
for more details. A demonstration
of the filtering mechanisms is available.
The COMET Group's research involves investigating the real-time behaviour of multimedia networks with QoS guarantees. They are involved in experimenting with architectural concepts that provide QoS to network applications and enable efficient use of network resources. Their paper, A Review of QoS Architectures, examines the state-of-the-art in the development of QoS architectures for distributed multimedia systems. QoS architectures are an attempt to unify QoS work at individual architectural layers to achieve end-to-end QoS guarantees. Current research in QoS support is discussed and 10 distinct QoS architectures in multimedia communications are evaluated. Joint work with Lancaster University examines adapting multimedia applications to fluctuating QoS. Details can be found in their paper Supporting Adaptive Flows in a Quality of Service Architecture. Incorporating QoS monitoring and adaptation mechanisms within an architecture that provides end-to-end QoS is reported in their paper Meeting QOS Guarantees by End-to-End QoS Monitoring and Adaptation.
The objective of the GRM project is to develop a framework for QoS-driven adaptive resource management in distributed command and control systems. System resources are allocated and controlled in accordance with user QoS requirements. QoS requirements are expressed as benefit functions which specify the benefit to the user or the system of attaining a given QoS level. Development is being done in the context of resource-intensive multimedia command, control, communications, computers, and intelligence (C4I) applications.
Meeting and sustaining QoS requirements requires monitoring and predicting the load on networked resources. The NWS, currently being prototyped, plans on using sensors to dynamically provide information on the availability and usage of network resources.
This research addresses customization of network services by setting up processes in the end-points of a cell-switched network to meet application QoS requirements. The full-feedback teleoperation of a robotics system is used to verify the approach. A QoS broker manages resources at the end-points and coordinates resource management across the boundaries of multiple layers in a end-to-end protocol architecture. Configuration of the system to application needs is achieved by QoS negotiation resulting in one or more communication connections. An evaluation of the approach can be found in the paper Design, Implementation and experiences of the OMEGA Architecture.
QUASAR stands for QUAlity Specification and Adaptive Resource management for distributed multimedia systems. The project is investigating QoS specification, and its use in both adaptive and reservation-based resource management. Application-level QoS specifications are translated into QoS requests that are meaningful at the resource-management level. Research emphasis is on multimedia systems that dynamically adjust to variability in available resources. See the Ph.D. thesis Quality of Service Specification for Resource Management in Multimedia Systems for details. This project is part of the research affiliated with Rowena, a latency-hiding storage system for multimedia applications.
BBN's QuO project is an attempt to extend QoS to distributed, non-multimedia
applications coping with a dynamic environment using OMG's
CORBA. A resource management
and availability model is introduced at the CORBA layer that exploits the
underlying QoS of the communication layer, and allows the actual QoS received
to be visible to the application thereby enabling it to adapt to changing
conditions. CORBA's Interface Description Language (IDL) is extended with
a QoS Description Language (QDL). QDL specifies an applications expected
usage patterns and QoS requirements for connection to an object, and change-of-state
handlers. The usage patterns specify the traffic constraints the client
agrees to abide by and the QoS requirements specify the QoS that will be
required when connected to an object. An application is allowed to have
multiple connections to the same object, each with its own system properties.
The status of a QoS agreement for an object connection is maintained in
an object's system states. The application can adapt to changing QoS parameters,
e.g., changing resource availability or partial system failures, by varying
its behavior based on the system state of its object connections. Such
an object-oriented QoS simplifies QoS management; an object is responsible
for managing its QoS instead of an external agent coordinating agreements
with multiple components.
The initial prototype is being developed for collaborative planning applications.
For an overview of this work, see the papers Overview
of Quality of Service for Objects and "Architectural Support for
Quality of Service for CORBA Objects" to be published in the January
1997 issue of Theory and Practice of Object Systems (special issue on the
OMG and CORBA).
Previously known as the Fast Web Project, the SuperNOVA project's objectives are to investigate QoS for the WWW, define an architecture which enables the real-time delivery of continuous media in the WWW, and specify a QoS negotiation protocol and a QoS adaptation scheme. See the Fast Web Project summary for details on the extensions to the Web architecture to incorporate QoS which can also be found in an IEEE Multimedia Newsletter which gives an overview of all distributed multimedia research at DSTC.
The main focus of the research group is on the design and development of real-time communication services and on network support for continuous media applications. An overview of the group's projects describes their R&D. A smoothing schema, described in the paper Improving QoS through Traffic Smoothing, allows network clients to optimize their QoS.
This research is sponsored by the Defense Advanced Research Projects Agency and managed by the U.S. Army Research Laboratory under contract DAAL01-95-C-0112. The views and conclusions contained in this document are those of the authors and should not be interpreted as necessarily representing the official policies, either expressed or implied of the Defense Advanced Research Projects Agency, U.S. Army Research Laboratory, or the United States Government.
© Copyright 1996, 1997 Object Services and Consulting, Inc. Permission is granted to copy this document provided this copyright statement is retained in all copies. Disclaimer: OBJS does not warrant the accuracy or completeness of the information on this page.
This page was written by Gil Hansen. Send questions and comments about this page to gil@objs.com.
Last updated: 1/20/97 gjh
Back to Internet Tool Survey - Back to OBJS homepage