34.3 QUALITY OF SERVICE QOS AND NETWORK PERFORMANCE NP You may have realized in our example above that the percentage of answered calls is dependent not only on network congestion, but
Trang 1Quality of Service ( Q O S ) and Network Performance ( N P )
The maintenance of good quality for any product or service (i.e its ‘fitness for purpose’ and its price) is of supreme importance to the consumer and therefore requires utmost management attention However, although it is easy enough to test a tangible product to destruction, measure- ment of the quality of a service is more difficult In telecommunication the customer is left with nothing more tangible than his or her own perception of how well the communication went On a datalink the errors that have had to be corrected automatically are barely appreciated, while in conversation unobtrusive bursts of line noise may go unnoticed This is not to say that loss of data throughput on a datalink caused by continual error correction is of no concern to the customer, nor does it mean that noises which disturb conversation are acceptable What we really mean is that in measuring service quality, due regard must be paid to anything of importance that the customer perceives and remembers Concentration on setting and meeting quality targets
should be paramount in planning and administration Insufficient attention to them is the road to customer dissatisfaction and loss of business This chapter reviews some aspects of communica- tions quality and the practices of management, with examples of the commoner quality
parameters and control measures used by the world’s major operators
Good management of whatever industry demands the use of simple, structured and effective monitoring tools and control procedures to maintain the efficiency of the internal business processes and the quality of the output When all is running smoothly
a minimum of effect should be required However, to be able quickly to correct defects
or cope with abnormal circumstances, measurable means of reporting faults o r excep-
tions and rapid procedures for identifying actionable tasks are required The framework for doing so needs to be structured and comprehensive In Table 34.1 a number of simple management ‘dimensions’ together with possible management performance tools/parameters within each dimension are presented As you will note the dimensions
cover a range of areas, some requiring more tangible monitoring measures and control procedures than others
633
Networks and Telecommunications: Design and Operation, Second Edition.
Martin P Clark Copyright © 1991, 1997 John Wiley & Sons Ltd ISBNs: 0-471-97346-7 (Hardback); 0-470-84158-3 (Electronic)
Trang 2634 QUALITY OF SERVICE (QOS) AND NETWORK PERFORMANCE (NP)
Table 34.1 Framework for performance management
Formal plan and delivery cycle Formal presentation and agreement of company strategy, (organizational framework) policy, guidance and plans
Regular review
Framework for business ‘value’ assessment and assurance User support
(fit for purpose)
Supplier performance
(quality of supply)
Quality performance
(effectiveness)
Financial performance
(efficiency)
Technical performance
(efficiency)
Resource management
(efficiency)
Evolution of networks
(responsiveness)
User comprehension of services (by questionnaire) Swift elimination of problems (full complaint log)
Maintaining professional attitude with suppliers
Monitoring and demanding swift lead times
Meeting targets for repair performance
Meeting user needs (number of complaints)
Number of lost messages (e.g percentage of data lost) Percentage of ‘down-time’ during normal business hours Measure of delay (e.g propagation delay, or backlog of orders/messages)
Measure of congestion (e.g percentage calls lost - callers frustrated)
Repair performance (e.g percentage not repaired in target time)
Connection quality (e.g percentage customers satisfied)
‘Cost of poor quality’ (i.e that correcting avoidable faults)
Return on capital investment
Shareholders financial return
Cost actually spent can be compared with competitors or alternative suppliers
Internal chargeout rate for telecom services (e.g ‘pounds per bit-mile’)
Compatibility of networks can be adjudged by measuring the money spent on interworking or upgrading and comparing it with total system value
Asset inventory and management (needs to be accurate and up to date)
Cost per task
Adequate resourcing to meet target service lead times Head count and man-hours should be maintained on target
Network upgrades should be properly planned and meet their budget
There should be steps to give ‘benefits’ assurance
The response time of new software and computer networks should be according to plan
The networks should be adaptable
Trang 3QUALITY: A MAREKTING VIEW 635
To look at a network from the customer’s viewpoint we need to understand his reasons for using telecommunications services It is a mistake to generalize about all customers
as ‘wanting to convey information over long distances’, as that takes no account of the service in use, or of each customer’s individual purpose or application It is, after all, a
Figure 34.1 A customer’s perception of quality Get the basics right before worrying with the frills Drawing by Patrick Wright (Courtesy of M P Clurk)
Trang 4636 QUALITY OF SERVICE (QOS) AND NETWORK PERFORMANCE (NP)
fundamental of marketing that products should be attuned to the markets and
customers they are intended to serve Thus the motivations and interests of someone who is out for the evening in calling home from a payphone, are quite different from the needs of a large multi-national company wishing to convey vast volumes of computer data around the world on sophisticated and dedicated networks at any hour of the day
or night
It may be that the same network infrastructure can accommodate the service needs of
a number of different markets or customer ‘clubs’ but it must do so without compromise One example of a network capable of supporting a number of services is the public telephone network, and it must run at optimum quality not only for human conver- sation, but also for facsimile machine interconnection A more advanced example of
integrated services networks is the integrated services digital network (ZSDN) discussed
in Chapter 10 and the B-ZSDN of Chapter 25
Customers’ communication needs can often be met in a number of alternative fashions, ranging from travelling in person, using the postal service, the telephone, telex, packet switching, and high speed data telecommunications services Pitching a given service to meet a given need is the job of telecommunications marketing specialists who determine its relevant qualities, expressed as a number of measurable parameters An example of such a ‘quality parameter’, used commonly by telephone network operators is the
percentage of calls completed Not only is this a fair measure of network congestion, it is also a good predictor of the level of customer frustration
34.3 QUALITY OF SERVICE (QOS) AND
NETWORK PERFORMANCE (NP)
You may have realized in our example above that the percentage of answered calls is dependent not only on network congestion, but also on the availability of somebody at
the destination end to answer the telephone Thus the quality of service enjoyed by the
customer depends on a number of factors in addition to the performance of the net- work ITU-T has also drawn this distinction, and its general recommendations on the quality of telecommunications services recognize two separate categories of perfor- mance measurement
0 quality of service (QOS)
0 network performance (NP)
Quality of service measurements help a telecommunications service or network pro-
vider to gauge customers’ perceptions of the service Network performance parameters,
on the other hand, are direct measurements of the performance of the network, in
isolation from customer and terminating equipment effects Thus quality of service encompasses a wider domain than network performance, so that it is possible to have a case of poor overall quality of service even though the network performance may be excellent The relationship of quality of service to network performance is shown in
Figure 34.2
Trang 5QUALITY OF SERVICE (QOS) AND NETWORK PERFORMANCE (NP) 637
quality of service (QOS)
W
network performance (NP)
Figure 34.2 The relationship between quality of service (QOS) and network performance (NP)
Similar parameters may be used to measure both quality of service and network per- formance (e.g propagation delay, bit error ratio ( B E R ) , % congestion, etc.) Normally
the measured quality of service is lower than the measured network performance The
difference is due to the performance degradations caused by the user’s own end equipment (Figure 34.2)
The measured quality of service will differ greatly from the measured network perfor- mance values where a connection is composed of a number of connections, traversing several different networks and end user equipments Although it is of utmost importance
to the end user, the problem with quality of service as a performance measure is that it is difficult to measure, and would need to be measured for each individual customer separately This is the reason for the development of the concept of networkperformance Network performance can be more easily measured within the network, and provides for meaningful performance targets for the technicians and network managers oper- ating the network
In short, QOS parameters are user-oriented, and provide useful input to the network design process, but they are not necessarily easy to translate into meaningful technical specifications for the network Network performance parameters, on the other hand,
provide a directly usable technical basis for network designers and operational
managers, but may not be meaningful to end users Sometimes, the same parameters are appropriate for both QOS and NP, but this is not always the case
The distinction between quality of service ( Q O S ) and network performance ( N P ) is
somewhat artificial, but it has become necessary as the result of recent regulatory
changes in some countries, where the public telephone operator ( P T O ) is allowed to
provide service only up to a socket in the customer’s premises, and end-user equipment
(i.e terminal) manufacturers slug it out in the customer premises equipment ( C P E )
market Government regulation should keep PTO network performance up to the
mark, but overall service may be of poor quality if it is let down by faulty or badly designed customer apparatus
Parameters should be chosen to reflect high quality end-to-end service and should
be expressed quantitatively as far as possible These QOS parameters should then be
correlated with one or a number of directly related network performance (NP)
Trang 6638 QUALITY OF SERVICE (QOS) AND NETWORK PERFORMANCE (NP)
parameters, each NP parameter reflecting the performance of a component part of the network, and therefore contributing to the end-to-end quality In this way quality of service problems can be traced quickly to their root cause
In the example of Figure 34.3, QOS criteria have laid it down that the signal loud- ness received by the listener should not be less than 14dB below the original signal transmitted The end-to-end QOS parameter is therefore signal loudness and is measured in dB The network between the two people in conversation comprises two
telephone exchanges and three linking circuits, which are designed to individual network
performance loudness losses of 1dB The overall network performance is therefore
3 + 1 + 3 + 1 + 3 = l l dB, and when the losses in the telephone handset, the end-to-end
(QOS) loudness is 13 dB, i.e giving a 1 dB margin within the maximum allowed loss of
14 dB
Considering only the network performance ( N P ) between the two telephone sockets
of Figure 34.3, an acceptable end-to-end QOS is delivered even if one of the links or exchanges in the network degrades in performance by up to 1 dB (provided that all the other parts continue to work at their designed losses) In other words, a network performance of 12dB is sufficient to deliver an acceptable end-to-end QOS of 14dB loss Likewise, if either of the customer telephone sets degrades in performance to the extent
of adding up to ldB, then the overall QOS of 14dB is still maintained, provided other components remain stable
In normal operation it is difficult to monitor the end-to-end QOS performance of all connections, particularly when a large number of permutated switched connections are possible, and the concept of network performance parameters is invaluable Each element of the network may be monitored in isolation and maintained within a pre- determined network performance range Thus in the example of Figure 34.3 we might aim to keep the individual exchange losses below 1.5 dB, and the individual link losses below 3.5dB Although theoretically this might mean that the overall network performance could include a loss of 3.5 + 1.5 + 3.5 + 1.5 + 3.5 = 13.5dB, which is 1.5 dB outside the maximum allowed network performance of 12 dB, it is highly unlikely that all the links and exchanges will simultaneously be performing in the most adverse manner As a safeguard against such an adverse occurrence, small occasional samples of the end-to-end quality of some representative connections will need to be taken Sometimes the relationship between the end-to-end quality of service parameter measurements and the network performance parameters may not be as direct as in the example just discussed In that example we were able to use similar parameters, all measured using the same scientific units (in our case dB, or decibels) However, in the
QOS parameter percentage of calls reaching answer, there is a much more complex relationship with measurable network performance parameters, depending not only on the state of network congestion, but also on the availability of alternative routes, the incidence of faults, and other abnormalities such as misrouting or misinsertion Some of the contributions may have negligible effect under normal conditions, so that reliable approximations may be used For example, if the percentage of calls being answered at
a given destination starts to deteriorate then the network operator’s first action is to look for any network links that are congested If any are found, then these are more than likely the cause of the congestion, and extra resources should be provided accordingly If no individual link or exchange congestion is immediately apparent then the operator will have to look for less common causes and faults
Trang 7QUALITY OF SERVICE (QOS) AND NETWORK PERFORMANCE (NP) 639
m
v
c
m
W
*)
1 - - -
m
V
m
a
2
Trang 8640 QUALITY OF SERVICE (QOS) AND NETWORK PERFORMANCE (NP)
There are no hard-and-fast rules as to which parameters should be used to measure quality of service and network performance, although ITU-T has created a generic
model and a set of generic parameters
Quality of service parameters focus on the user’s likely perception of the service, rather than on the technical cause of specific degradations of service For this reason, different quality of service parameters should be tailored for each different type of network application
QOS parameters are not comprehensively defined by ITU-T standards For certain types of services (e.g leaselines, telephone service, etc.), ITU-T defines QOS parameters, and these same parameters should be used to measure end-to-end quality of such services, independent of what type of transport network is used Thus, for example, the quality of service of a digital leaseline could be measured in terms of the bit error ratio ( B E R ) , the jitter, the accuracy of the bitrate, the propagation delay, the availability of
the connection (the percentage of a one-month or three month period for which the circuit was not out-of-service) and the accuracy of the invoice
In addition, the network operator has an interest to conceive network performance parameters which would help him to meet the user’s expected QOS level In this respect,
he has a little help, as ITU-T has defined a number of standard network performance parameters for some types of networks The network operator’s task thus becomes relating these to the specific expectations of individual services’ QOS, and setting appropriate network performance targets
34.5 GENERIC NETWORK PERFORMANCE PARAMETERS
For connection-oriented networks (such as the telephone network, packet and frame- switched data networks and ATM networks), ITU-T defines a set of three different types of NP parameters, and recommends that these should be used as measures of three different functional aspects of the connection (access (connection set-up), user information transfer and disengagement) Table 34.2 shows the nine different types of
Table 34.2 The generic primary performance parameters Performance
Access Access speed Access accuracy Access dependability User information Information transfer Information transfer Information transfer
Disengagement Disengagement speed Disengagement Disengagement
accuracy dependability
Trang 9GENERIC NETWORK PERFORMANCE PARAMETERS 641
Table 34.3 Example primary performance parameters relevant to telephone, ISDN, data and
ATM networks and services
Performance
Access Connection set-up
delay
User information Successful transfer
transfer rate
Propagation delay Cell transfer delay Cell delay variation (CDV)
Cell transfer capacity Jitter
Disengagement Delay in connection
clearing
Incorrect set-up probability (misrouted connection ratio)
Bit error rate (BER) Packet, frame or cell misinsertion rate Packet, frame or cell
error ratio (CER) Severely errored frame or cell block ratio (SECBR) Errored seconds Premature release
ratio
Incorrect release ratio
(CMR)
Probability of set-up denial (connection set-up denial ratio)
Probability of information loss Packet, frame or cell
loss ratio (CLR)
Release failure ratio
network performance parameter which result These are termed the generic primary
performance parameters Similar performance parameters could also be conceived for
connectionless networks
Examples of specific primary performance parameters within each of the generic per- formance parameter classes are given in Table 34.3 A subset (perhaps all) of these para-
meters should be regularly measured by network operators (as network performance measures) The target values for the chosen parameters need to be set to ensure relevant customer satisfaction with respect to quality of service Where no direct relationship exists between quality of service and network performance parameters, experience will help to determine acceptable threshold values for NP parameters
In addition to primary performance parameters, ITU-T also defines the concept of derived performance parameters The most important derived performance parameters
are those of availability and acceptability Availability is a measure of the cumulative
outage (i.e non-service) time of the network as a whole (or of a given customer’s part of the network) during a given measurement period (e.g one month or three months) Availability is measured as the percentage of the total period for which the service was not in outage Thus the higher the measured availability, the lower the network outage Typical target values are 99%, 99.5% and 99.9%
Acceptability has also been proposed as a derived performance parameter This would be intended to give a qualitative measure of likely subjective customer opinion of the service level This gives the potential for inclusion of other more general factors in measuring overall customer satisfaction with the network
Trang 10642 QUALITY OF SERVICE (QOS) AND NETWORK PERFORMANCE (NP)
34.6 PERFORMANCE MONITORING FUNCTIONS OF
MODERN NETWORKS
In the most modern network technologies (e.g SDH, synchronous digital hierarchy (Chapter 13) and ATM, asynchronous transfer mode (Chapter 26)), extended perfor-
mance measurement and monitoring techniques are built-in Thus, for example, it may
be possible to measure the instantaneous bit error ratio ( B E R ) of a given connection, or
the propagation delays being experienced in a live network This is done by inserting extra performance management traffic into the network and monitoring the experienced performance This is intended to be a good predictor of likely network performance as perceived by end customers
The performance monitoring function of ATM operates by sending performance monitoring cells on an end-to-end basis after each block of N user cells This function
can be used to monitor end-to-end errored blocks, misinsertion (incorrectly sequenced or
incorrectly delivered cells), cell tranqfer delay and other performance parameters on a
specific connection
As an example of the network monitoring and diagnosis tools available within ATM,
it is, for example, possible to apply a loopback to O A M (operations, administration and
management) cells connections within an ATM network without affecting other virtual connections sharing the same physical connection This may be useful to confirm circuit continuity and to measure loop propagation times, without having to interrupt the live user connections
34.7 NETWORK PERFORMANCE PLANNING AND MEASUREMENT
The remainder of the chapter proposes a pragmatic approach for establishing QOS and
NP parameters for telecommunications services The approach is designed to generate a balanced mix of parameters that will prove valuable in monitoring all aspects of service quality, including not only the quality of the connections themselves, but also the
important support activities such as maintenance and provision of service
By defining a standard set of categories relating to telecommunications service quality,
we can ensure that selected parameters within each category give a balanced picture of current network performance and quality of service as against target Without such discipline in the past, many network operators have made extensive measurements of the technical aspects of performance with too little respect for the broader service attributes
giving them a distorted view of their quality, and they have suffered accordingly For although they may have been returning exemplary technical performance statistics, their customers may have been far from happy because they could never get a reply from the fault reporting bureau, or could never get assistance about how to use a service A pos- sible set of categories from which to choose a balanced set of performance parameters is shown below Each category is explained, and some examples of actual parameters are given in the paragraphs that follow
0 customer service
service availability