With the development of high technology network, more and more people are starting to use networks. At the sametime, more and more businesses are involved in using network. How do they choose between different networks and services? QoS-Quality of service parameters play an important role in making these decisions.
QoS can be defined in several parameters. Those parameters can be used for evaluating different networks. According to those QoS parameters, a user should easily find what kind network service he can get. If the current network services can not provide enough relevant QoS parameters the user demands, then what should we do? Obviously, we could first find out what is really missing form current network QoS by considering user's demands.
In order to give a network user some ideas about QoS of current networks, in the first part of this essay we will summarize the main QoS parameters for different networks. In the second part, we will consider QoS parameters which a user demands. In the third part, we will discuss what QoS parameters are missing from current network services.
There are three QoS parameters with relevant to LAN. They are:
The desired basic parameters for the connection:
By considering each different QoS parameters, a user can easily choose a special network service which meet his demands. For example, if a user demands constant data transferring rate, constant delay, he could choose a CSPDN ( circuit switching public data network). If the user wants to transmit a large amount of data with loose time limit, probably he will choose datagram transmission --one kind of packet-switching PDN which is much cheaper than a virtual connection and CSPDN. If the user wants to choose a specified route, he can choose a permanent virtual connection.
As we see, QoS may vary from one network to another. When a user is using internet, there are much more choices available than when using a single LAN or a special WAN. QoS is an even more important fact for user choosing proper networks. There are 8 QoS parameters that user expects from the provider.
In comparison with a WAN, three more QoS parameters are added in the internet. Congestion control specifies whether flow control will be exercised by the internet. Sequence preservation probability is the result of a measurement which indicates the ratio of sequence-preserved transmissions to total transmissions to total transmissions. Maximum NSDU lifetime indicates the maximum time the internet is allowed to take to deliver the NSDU before discarding it. This parameter gives the user some idea about maximum time he should wait for receipt of acknowledgment before he can transmit any data.
There are five classes of service which are defined by ATM. These five classes of service are based on following three criteria:
Based on above three criteria (six parameters), following five classes of service are provided by ATM.
In a ATM network, class A provides highest quality of service while class E provides lowest quality of service. It is easily to see that to have time relationship and constant bit rate requires high level quality of service. The QoS associated with connection-oriented service is higher than with connectionless service. In the connectionlees service, data is transfered directly by using the N_UNIDATA service. In the connection-oriented service, a connection is first established before data can be transfered.
As a user, he pays more attention to what kind services a network provide could provide. Maybe he doesn't if he needs any networking service with flow control or if a virtual connection is built. But he does know how long time he should wait before he could transmit his data, how long time receiver could receiver his message, after that, how long time interval does he want between received messages, will he receives a error-free message? How much is the error probability? Could he be sure he can always send data in any time? How much should he pay for the service? Could he get any flexibility in the service he gets? Could he directly controls QoS parameters he wants? Could transmitting processes protect any unauthorized monitoring or modification of data? All those above facts have more direct relation to users. We can summarize above facts in following QoS parameters:
Time relation( Yes or No ) and bit rate( Constant or Variable) parameters can specify data transfer delay. Flexibility of a network service means that a user can negociate with a network service provider . Except exact rules a user can get from a provider, he should able to discuss with the provider and tells the provider which is acceptable and which is not.[1]. Availability means that services are available when a user requestes it. The larger data transmission capacity and the lower failure rate a network has, the better availability it can provides. Security level is a parameter from which a user can know if the network service could provide any security protection for his transfered data.
In comparison with the QoS parameters a user demands and the present parameter a network provider provides, we can see flexibility, availability and security level are missing from present QoS parameters.
In a real world, the following facts can seriously affect QoS:
Although random hard ware failure happens rarely in a well designed and maintained network, when it happens it still can seriously damage a provider's quality of service image. Flow congestion is a theoretical problem with QoS of a packet switching network [2]. If a flow congestion happens, the network service can not guarantee any QoS parameters it provides to a user,although it happens very rarely.
In conclusion, QoS parameters are provided not only for us to judge the system but also for user choosing different network services. When we design a network or provide any services, user's demands should be considered first. Hence we could provide enough necessarily QoS parameters to users.
[1]
Jan-Peter Richter,
Flexibility in QoS Support,
4th CaberNet Radicals Workshop,
1997.
<
http://www.fcul.research.ec.org/cabernet/research/radicals/1997/papers/qos-richter.html >
[2]
Vassilios Tsaoussidis,
ASQoS: Application oriented System for Quality of Service,
1997.
<
http://www.caip.rutgers.edu/~vassilis/asqos.html>
[3]
Ingo Busse, Bernd Deffner, Henning Schulzrinne,
Dynamic QoS Control of Multimedia Applications based on RTP,
1995.
<
http://www.fokus.gmd.de/step/acontrol/ac.html>
[4]
J. Kemper,
ETH Zürich Mobile Radio Group research announcement
1997.
<
http://www.nari.ee.ethz.ch/nari_mobile/mobres.html >
[5]
Douglas C. Schmidt,
QoS for Distributed Object Computing Middleware -- Fact or Fiction?
Panel at the Fifth International Workshop on Quality of Service (IWQoS '97),
1997.
<
http://www.cs.wustl.edu/~schmidt/IWQOS-97.html>
[6]
Jakka Sairamesh,
Economic Paradigms for QoS in Information Networks,
Research announcement,
Columbia University, New York, NY,
1995.
<
http://www.cs.columbia.edu/~jakka/pricing.html>
[7]
T. Saito, H. Aida, T. Aoki, Y. Kishi, and P. Setthawong,
QoS Guarantees for High-speed Variable-length Packet LANs,
Faculty of Engineering, The University of Tokyo,
1997.
<
http://www.sail.t.u-tokyo.ac.jp/~pisai/research/canada/can.html>
[8]
William Ivancic,
ATM over Satellite Quality of Service Testing,
NASA Lewis Research Center, Cleveland Ohio
1996.
<
http://sulu.lerc.nasa.gov/wivancic/atmintro.html>
[9]
Gideon Stupp,
Application Network-Layer Quality of Service (QoS),
1996.
<
http://theta.math.tau.ac.il/QoSrmi/report/html/report.html >
[10]
Yehuda Afek, Michael Merritt, Gideon Stupp,
Remote Object Oriented Programming with Quality of Service orJava's RMI over ATM
1996.
<
http://theta.math.tau.ac.il/QoSrmi/mascot97/abstract/abstract.html >
[11]
Nalini Venkatasubramanian, Klara Nahrstedt,
An Integrated Metric for Video QoS,
ACM MM'97,
1997.
<
http://nahrstedt.cs.uiuc.edu:5000/~klara/ACM-MM97/paper.html >
[12]
Mambretti,
Multi-Queue Management and Scheduling for Improved QoS in Communication Networks,
European Multimedia Microprocessor Systems and Electronic Commerce (EMMSEC 97),
1997
<
http://www.anl.gov/MREN/workshop/mambretti/index.htm>
[13]
Manolis Katevenis, Dimitrios Serpanos, Evangelos Markatos,
Credit-Flow-Controlled ATM over HIC Links in the ASICCOM
"ATLAS I" Single-Chip Switch Real-Time Magazine vol. 96, no. 3,
1996.
<
http://www.ii.uib.no/~markatos/avg/html_papers/EMMSEC97/paper.html >
[14]
Jon Crowcroft,
QoS - Plenty of $$$ but just 1 Bit ,
3rd International Workshop on High Performance Protocol Architectures HIPPARCH '98,
1996.
<
http://www.cs.ucl.ac.uk/staff/jon/hipparch/dollarbit/index.htm>
[15]
Doug Montgomery,
ANTD Project Proposal: Support of Integrated Services in the Next Generation Internet Protocol Suite,
1995.
<
http://snad.ncsl.nist.gov/iieg/ispsn/IPS-IS/IPS-IS.html >