ADSTRACT The past few years witnessed a rapid growth of interest in network reliability. This is because telecommunications services are now an integral part of businesses, national security, Md public health Md safety. Also, recent technological advances such as fiber optics highcapacity digital switches and the increasing concentration of capacity in the telecommunications infrastructure have made networks more vulnerable to single failures I$ecause of this capacity concentration Md the increasing reliance on telecommunications the potential impact of a single outage bas never been greater on business public perception, Md geographic scope Du QSDG work attention bas hem paid to the network reliability and network outages In particular, the sud for developing a methodology for quantifying the customers' impact of a network outage has been recogsised Network reliability is one of the major future items of the QSDG. The Quality of Service Development Group ( QSDG ), which is in nature an operational group including a field trial group, was created in the Study Group 2 orgauisation in 1984. The original idea came from Mr. C. McCauley from AT&T who felt that, despite their usefulness, the writing of Recommendations as such or drafting a handbook was not enough to improve the quality of service Telecommunications people need to come together, communicate Md help each other, to achieve the best actions for improvements Md to have a free exchange of views Md ideas in a multilateral environment Thesupport was overwhelming Md fTC Study Group 2 has approved the establishment of this Group mainly because many Administrations Md RDA's felt that practical approaches to improving quality of service should he implemented Md this could best be done by a Group not burdened by mJ"T (then CCfTT) procedures where "formal agreements" must be reached on Recommendation. The first meeting of the QSDG was held in Hague The Netherlands in 1984 Md very soon the participation increased as cu be seen in the followieg table
YEAR
LOCATION 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995
The Hague
Orlando Hong-Kong
Manila Albufeira
Rem Padova
81085 S. Francisco
Bolero&
Chester Joio Pessoa , , ,"
The primary aim of the QSDG is to improve the qnality of the international telecommunications services to the motual benefit of both the costomers and the Administrations/Roe's The..QSDG in its Work Program will continoe to study all aspects of Qnality of Service from the point of view of the cost.mer who oses the telecommunications network The Group is conscious that in addition to the voice use of the PSTN, there is also a need for it to cover non-voice uses e g fax ISBN, R-ISBN, packet video and other services The work inclodes general studies on how costomer service cm be improved, inclnding the interpretation and ose of costomer surveys, how information cm be eachanged between Administrations/ROA's, the effect of network digitisatioo, appropriate network measurements and their correlation with customer input and multilateral benefits of improving qnality of service and network performance Considerable work has been accomplished, and several achievements bu been reached. An example is the "MD Completion Rates World-wide" database which is updated yearly and osed extensively by the participants From this database it bas been concluded that the number of destinations which coold be considered as offering poor call completion is decreasing Following the criteria deranedin the Recommendation E.426, the following table provides insight Lathe ASK ( Answer Semore Ratio ) trends
ASRTIE > 60% > 30o,o < 60% < 30 1991 9.8% 61.4% 288% 1993 14.6% 67.1% 18.3% 1994 18.9% 61.6% Considering the information in this database and taking into accoont that QSDG bas met 6 times in Enrope 3 times in Asia-Pacific Region, twice in USA, and once in Sooth America it was decided that fotore meetings should be held in regions of the world where QSDG had not met before It bas been felt that if QSDG meeting took place in these areas, the number of delegations participating would increase, and more people would be involved in Qnality matters Conseqoently it is esfaected that the quality and performance arose destinations in these areas will be improved. This objective has been achieved and the neat meetings of QSDG are planned to be held in Africa Middle East and in Oceania by the kind invitations of Sooth Africa Bahrain and Fiji Islands, respectively.
It is anticipated that as long as the participating entities are benefiting from it the QSDG will continne to exist and will meet around the world regularly. All costs involved are dome by the --.~~~ :. " \ , ^ -_ " No one denies that computing and telecommunications are converging, but the vision of the converged world has changed Until recently, it was still possible to imagine that the future would look pretty much like today's telecommunications environment bot with fancier tricks, like video built into the familiar telephone, Md maybe even telephone companies selfing combined PC-phases it is now becoming clear, however, that the reality win coasist~of more powerful Md flexible compoters that are capable of tavlag over most or maybe even all, of the traditional telecommunications foactioos Convergence increasingly means a competing tal;e over. This may still be controversial among some of the larger private branch exchange (PBX) vendors Md telecommunications service providers, bot coafldeace in the compuLtingcamp is growing to such an extent that some are eves handing oat diagrams showing the information technology business environment of the near future, Md there is no room in the diagrams for a PBX Today's CTI (computer telephony integration) systems generally Calfinto one of fame different architectures or configurations based on their approach to making the actual connections Md making calls: * Phonecentric systems * Serverceatric systems * Voice server systems * PCcentric systems Phone-centric sYama are the easiest to implement; they only require a direct link from the phone to an external adapter that connects to the PC's serial or paralld port They don't require extensive changes to an existing phone systems Users cu have direct control over calf routing (known as first-part call control). To transfer a call, for example, the nser Josi cHcks oa an icon, Md the PC sends a message to the switch that emulates a command from the phone reqoestWg the switch to trander the can. Many PBX vendon offer adapten that give that kind of control to the PC. Unfortunately, these adapters don't provide a connection to the phone line Md can't be used to connect data or fax lines to the PC.
Smercenhic sVaems connect your telephone switch to a server oa yoor LAff. Here the phone system becomes another part of the computer network Md you don't need a physical connection between the phone and individual desktop PCs The LAN server, not the switch or the user, is responsible for routing calls Ohos it's termed third-party caHcontrol). To transfer a caH the oser clicks on the transfer icon, which sends a message to the server requesting that it transfer the calf The third part (the server) sends a message telling the switch where to mote the call The server's processing power Jen it screes Md mote incoming calls For example, caller-m informadoa may help mute the call to the proper perms. Third-party call control is particularly helpful w workgroops and call centers nut the servercentric modd manages only calf control The switch-tnserver link is for status M d reqoest only. It doesn't carry the voice path and in no way physically connects a phone line to the server. For a server to send and receive faxes and data a physical phone line would have to be connected to a fax-modem hoard in the server. Voice-server sY5tems are a variation on the servicecentric model. Where servercentric systems deliver callcontrol links but not the calls themselves, voice server systems deliver the call directly, hut not a separate control-and-status link In a voice server model, phone lines from the switch connect to a board in the voice server. Depending on the board's capabilities the lines cm he analogue, ISBN, or proprietary digital The board cm do anything that a phone it replaces cm do; for example it cm issue a Nash hook to transfer, conference hold, call park call forward, initiate call pickup from another office, and so forth. Digital phones usually have other features, such as speaker-phone control and calfer-ID display. With the phone line going into a voice server, you get the media - that is, the voice path, or data path for faxes and modem calls - but you don't get all the information and control that's available on the serverceatric model. For example a phone line can't force the switch to take control of mother call; it cm only control a call that it has received or placed. It ran'f ten the switch to forward a calf from the next office to mother phone In a servercentric callcenter application, the server receives the railer-ID and tells the switch where to send the call. In the voice server model, on the other band, the call is sent to the server, which must then answer it and transfer the call. But this is just too slow for a call center. PC cmbw" systems have the telephone line and the telephone itself connected directly to an addin board in the PC. The telephony board emulates the type of telephone that the switch is designed to support, whether analogue or proprietary digital When we have isochronous Ethernet or ATM (asynchronous transfer mode) data pipes going directly into our PC& which looks to be the long-term prospect, we'll use PCcentric telephony systems For the short term, however, we'H see fax-modern boards with telephony features that will provide an interim solution.
Just as a crisis hit the mainframe computing world when PCs first burst onto the scene PBXs are seen as the inevitable victims of smarter PCs that cm understand human speech, play realtime video and still have enough spare random memory to mu telecommunications functions in softwarc Equip such a PC with some slot-in cards that can send and receive telephone signals, and computing bas effectively swallowed telephony.
There will stiH be space for the PBX vendors, just as there is still, even today, a place for 'mainframes in the corporate computing world. fa the office backbone network and in the wide area networks that links corporate offices convergence is taling the form of a new communications protocol that cm handle data voice and video with more or less equal ease Called asynchronous transfer mode (ATM), this protocol prowises to be the basis of new kinds of networks and way eventually - though this still controversial - extend alf the way from the international networks to the desktop-based local area networks ATM is stiH fertile ground for the PBX vendors whose detailed underenvironment wakes thew prime partners for the computing and data networking vendors The larger PBX vendors are also manufacturers of public telephone exchanges, and they have been among the front-runners in developing ATM technology .
The requirements of the public networks in the converged world are difficult to foresee however. A lot will depend on what happens at the individual easterner sites, both in home and in the omce- Much as the pubHc telecommunications carriers woald Hke to forecast the demand inthe next decade or even to shape that demand to suit themselves the chopandcbange nature of the computing world is hound to make that wish impossible to realise What is hardest to predict is the extent to which computer power wifl infiltrate the public exchanges sites. themselves It is already possible to build a pabHc telephone exchange out of standard computer components Some believe the resulting system cannot possibly be resiHeat enough to provide the service availabiHty that telecommunications customers expect but the reality is that computing is becoming increasingly more tenable Indeed, the problem of 100 percent availahHity has already been solved, time and time again, by computer systems daigned for banks rwance _ ~ ~ -~ _ ,_ ~--- , -compuies and the military. It used to he that all telecom systems required hardware and software more sophisticated that what was on offer for standard business computing solutions This is increasingly no longer the case Standard computers are now powerful enough to handle the human voice and fuH-motion video in real time The operating systems running on today's servers are sophisticated enough to cope with the logic required of a digital telecommunications switch.
In fact, the insides of the telecommunication boxes which used to look sophisticated compared with corporate computer systems now look out of date PBXs and telephone switches actually contain their own computers, operating systems and applications software Typically, the computers are non-standard, the operating systems are proprietary, and appiications are not flexible There is no real need for all that computing to be inside the telecommunications boz at all Transferring the guts of the PBX to a UNIX or Windows NT server suddenly opeus the market up to price competitive hardware to operating systems that integrate properly with corporate information technology and tosoftware that cu be modified by the customer. In the long term, the same will probably he true of the public telephone exchange What this means for the world-wide compatibility of systems and services is an important question. We are growing accustomed to being able to pick up a telephone in one part of the world and get through to my other telephone on earth. The more complicated and varied the systems get - and more they come to resemble computers - the harder it will be to maintain its unfformify.
In 10 years, convergence will have changed the ink and feel of telecommunications and computer technology, and also the shape of the information technology industry. Just as the advent of distributed systems revoiutionised the industry in the 19808, convergence will do the same between now and the miHennium- The previous revolution forced some companies to merge especially the larger system providers, while some once obscure companies rose to international prominence The convergence revolution is bound to make a few reputation, but it may also break a few.
4 TL. nCfb n ml [Lo -a7;nA;7;?t;zv icemf a .
The use of International Telecommunications services is dependent upon the interconnection of numerous items and media which are controlled, often independently, by various entities and agencies This is increasingly true also for national teiecommunications services as deregulation proliferates WhHstindividual administrations, agencies or network operator may be subject to desirable Quality of Service measurements and improvement programs, it is the overali (end-to end) performance which customers perceive as important When creditcard, longdistance telephone and retail banking services users are asked to rate (on a scale I to 10) the importance of the different dimensions of the Quality of Service reRabiiity clearly emerges as the most important dimension, regardless of the service being studied. A typical example can be seen in the next table Service Quality
Dimension
Responsiveness Assurance Empathy
Card Average Rate 7.43 9.45 9.37 Gd _ ---, .~ -~ --_
The customer's message to the service provider is clear: Be responsive be reassuring, be empathetic, and most of all be reliable Complementary there is another message: Human performance plays a major role in customers' perception of service quality. Three of five dimensions - responsiveness assurance, and empathy - result directly from human performance Moreover, reliability often depends largely on damn performance Concerning this particular point, it has been felt that QSDG, in their working methodology, is playing a significant contribute The reliability performance of a system is its ability to perform a require function under given conditions for a given interval. Important factors influencing reliability performance are: The quality of the constituent components Environment factors The network structure
Security against secondary failures Telecommunications networks are vulnerable to many threats: natural disasters, intentional sabotage, accidents, and faulty hardware and software But technical solutions exist and cm be deployed to minimise both failures in network elements and the impact of disruptions on telecommunications infrastrucWres, thereby meeting the requirements of telecommunications users Telecommunications services have traditionally bun cbaracterised by high level of service dependability. Dependability is a characteristic that includes reliability, availability, maintainability and survivability. In briefly, dependability means that a service works the way a customer wants it to work when the customer wants it Three different considerations are making service dependability a topic of great interest in the industry today. First as the use of telecommunications services bu become a more integral part of many businesses, customer expectations for dependability have increased Second, revolutionary changes are being made or proposed in the network While each of these changes brings with it the opportunity to provide customers with additional services that nut their muds in a timely fashion, each also provides a challenge in assuring the continued dependability of new services Third, some unfortunate events in the recent past have disrupted services for a large number of customers for m extended period of tine and have caused concern among customers It is recognised that a three pronged attack is needed to meet the challenge at band: * M31 lug the network elements less vulnerable to hardware failures, software errors, and procedural errors, and therefore minimising network elements outages * Mlling the network less vulnerable to network element outages, for example, through use protection switching, diverse routing for protection, scffhealing rings, dual homing, and other mens of minimising network outages or theimpact of the outages o Improving and automating the techniques used to restore service after a network outage Beliable network elements, survivable network architectures, and efficient restoration strategies will be the keys to assuring dependable telecommunications services in the 1990s Networks have traditionally not bun designed to provide 100% service continuity in race of catastrophic failures because the cost was thought to be too large However, with the increasing dependence of customers on telecommunications for their business survival, customers expect a high level of service restoration after a catastrophic failore More and more, costomers require and demand service approaching 100% availability on an end-tocad basis Even if customers settle for less than 100% availability during a failore, their minimal reqnirements may be 100o/o availability for critical services, with some minimum level of performance for all services to aft locations Network providers must balance meeting requirements of specific customer and service with providing a high-reliability network with service assurance to all customers The ultimate reliability goal is to make all failnres imperceptible to users An interim goal may be to rednce the impact of a failure so that calls in progress are not cutoff and data sessions are not prematurely terminated The past few years witnessed a rapid growth of interest in network reliability. This is because telecommunications services are now an integral part of businesses, national security, and public health and safety. In addition because telecommunications services have traditionally been so reliable even through national disasters, public expectations are very high Also, recent technological advances such as fiber optics highcapacity digital switches and the increasing concentration of capacity in the telecommunications infrastructure have made networks more vulnerable to single failores Becaose of this capacity concentration Md the increasing reliance on telecommonications, the potential impact of a single outage has never been greater on bnsiness, public perception, Md geographic scope Therefore, there bas been a concerted effort by endusers, service providers, standards bodies, and government and regulatory agencies to undertake multiple activities addressing issues related to network reliability and service Dotages over the past few years This will lead to a scientific assessment and enhancement of network reliabuity. A major question is bow to quantify a network outage For years network reliability planning bas concentrated on establishing the reliability of each network component The individual component reliability are then aggregated to establish, through modelling, a target network reliability. This resulting reliability was then examined for adequacy.
How does a network planner determine what is adequate? The arrest view of quality suggest the planning start with customers expectations and needs for their services Those service reliability needs cm then dictate the demands on the network and in turn the individual network component requirements With this current methodology, then, the onderstanding of reliability levels delivered to customers must also begin with an end-toend network assessment of what quality of reliability is being delivered to the customer. Thus the process now begins with the customer. Reliability as defined by AGREE (1957) and ANSUASQC is the probability that an item will perform without failure a required function under stated conditions for a stated period of time To define network reliability, QSDG bas accepted to describe the six elements of the reliability definition for a telecommunication network In network applications the six elements of the reliability dermition are: =>Probabitity - Network reliability is quantified in terms of probability. =>Item - The item to be considered here is a telecommunication network However, a variety of distina, but interconnected, networks exist These networks could be owned and operated by different companies For instance, there are multiple LMal Exchange Carriers (I.ECs) and Interexcbange Carriers (ICs). A single end-toend usage of the network may utifise several of these distinct interconnected networks fn network reliability studies the network under investigation needs to be precisely defined Md its boundaries specified. _ . ~ -~ ~ -._ --~_ _', -^ ~ =>Required function - A telecommunication network is required to perform several function& From the users' perspective and the network reliabuity point of view, the first and the most commoa fuactioa of a aetwork, is the abHityto communicate from a source to a target This is the ability to establish a sew connection and/or maintain the established connection. =>Failure- A network fails to perform the required fuactiou, from a source to a target, whee the failure probability exceeds a pre specified failure threshold value (the maximum designed value). The failure threshold and/or customer impactiag values Reed to be determiaed for differeat aetworks, applicatioas and services =>Stated coaditious - A telecommuaicatioa Remark consists of many distiact elements Networks are designed to perform the coanectioa establish.eat fuactioa under certain coaditions These coaditioas depend on the aetwork, service, and applicatioa. QSDG is considering that a failure due to CPR (Customer Premises Equipmeat) should not be coasted as a Remark failure =>Stated Period of Time - A time period must be specified in order for a probability measure to be meaniagful. It is desirable to know what is the probability a aetwork will perform the connectioa establishment fuactioa duriag the next time interval.
It is seen, therefore, that, a statement about network reliability requires an explicit, dearly defined and formulated descriptioa of the Remark under sway,the Remark functioa, aetwork failure and allowed eavironmeat and couditioas Finally, a network outage needs to he clearly denned in terms of customer perceivable and measurable quantities The customer's perspective of a Remark outage can he dassified into three major components of Inteasity, Duration, and Extent The (IDE)-triple, provides a framework for measuring network outage This framework cm be applied to differeat networks and services Each Remark and/or service needs its own specific formulae and computatioaal procedures Quantitative scales measuriag end-toead customers' impact of an outage may be devdoped through appropriate inWgratioa of I D, and E, as well as customers' relaWd parameters The Federal Communicatioa Commission (FCC) of the USA issued a report (February 27, 1992) defining "service outage" to coastitute loss of service to more than 50000 customers (later reduced to 30000) for a period of 30 or more minute& Many international organisatious fearful of a damagiag loss of customer coafideace refuse publicly to admit ever baviag suffered a major aetwork outage and are reticeat even to acknowledge the possihuity. New Jersey-based Ascom Timeplex, is oae of the few companies to have a successfully commissioaed indepeadent research. Its survey of 100 of the top companies in the UK shows that 50% of those iaterviewed predict a loss of busiaess if a communicatioas network is down for more than an hour. Almost alee out of tea firms (87%) said a aetwork f~ailureof less than 60 minutes would have a measurable impact oa efficieacy and throughput A similar survey iadicates that 73% of respondeats experienced IAN failures, 23% of which lasted for more than 10 hour&Some companies logged as many as 20 outages a year. Failure was particularly high oa aetworks supporting between 100 and 299 users, with router and bridge iaterconaectioa technology beiag the most likely to go down. Furthermore a high degree of the recurrence of specific problems was reported. Token riag networks appareatly failed more often than Ethernet, but disabled fewer users whee it was out of Reties. Research specific to the US oaly shows that a typical LAN there is likely to be disabled twice as oftea for an average of five hours These fiadings are bolstered by a European-wide survey of 100 top IT professioaals, conducted by the Yankee Group. Dee 6nancial instiW6on is on record n sayieg that it could easily lose US$l0000 in direct costs for every minute that its aetwork is down. Indirect costs, such as loss of credibuity and market share breach of statutory requiremeats casbflow problems and collapse of share price make the overall expense of major ootage, literally, incalculable Within QSDG, work on network reliability started on 1993. Network reliabuity is one of the major future work items of the QSDG. In particnlar, the meedfor developing a methodology for quantifying the customers' impact of a network outage bas been recognised by regulatory agencies industry leaden and end-osers A network reliability policy begins with identification and defioition of quantitative scales to measure an outage and assess its impact on endusers