Your SlideShare is downloading. ×
White paper0
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

White paper0

442

Published on

Future Trends in Telecommunications Network Planning

Future Trends in Telecommunications Network Planning

Published in: Art & Photos, Business, Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
442
On Slideshare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
2
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. White PaperNew Universal Evolution of Telecommunications Network Planning: Fixed-Mobile Convergence with Application and ServiceDivergence Prof.Dr.Dipl.Ing.Mehmet Erdas SIEMENS PSE TN MNS SA mehmet.erdas@siemens.atAbstract:The unification of both packet- and circuit switched world leading to the convergence of fixedand mobile networks together with the database planning, necessitated the concomitantintroduction of new applications and services in order to protect the previoustelecommunications investments of operators or investors, who are urgently looking for theprotection of their investments. This paper gives an overview of how to optimize the networkplanning problem by maximizing benefits while minimizing the risks of the investors-suppliers-and operators, under the light of latest developments in the area of networkplanning like the self-similarity of traffic, dynamic routing and topological constraints ,summarizing the scope and the methodology of network planning and network management.The next generation of access, switching, and transmission networks, as well as the end-user IT-equipment will be much more faster and intelligent, much more self-contained-andactively self-regulating, adapting themselves through their own iterative adjustment anddecision making mechanisms to their own conditions or adaptive goal-settings. .The New Network Planning and Network Management Teams has to act at the speed of lightor at least at the speed of thought, in order to be able to keep pace with the accelerated rateof change of the market demand (All-IP, IPsec, DiffServ QoS),as the main driving force of thetechnological innovation.The new stored program controlled complex procedures with fullyopen interfaces will make the active networks soon a market reality leading to theconvergence of the fixed-and mobile networks. The All-IP (Ipv4 replaced soon by Ipv6)convergence of different protocol stacks are to be achieved by the unification of circuit-andpacket switched technologies through queueing theory and stochastic discrete eventsimulation.The new unified network architectures enabling gigabits or terabits of throughput withunderlying topological configurations and dynamic routing policies will in themselves beoffered as a totally new revenue generating business, service or product. The world of ISPsor ASPs will be too complicated facing the question of survival against rapidly changingchallenging market conditions. The Operators –Suppliers-and Investors of futuretelecommunications products or services will definitely need the network as an intelligentunique product in itself offered by the Network Planners or Consultants, because ISPs /ASPsare just about to loose their own control, or having to outsource , at least losing soon theirdegrees of freedom and/or competence in decision making to the network planners of theuniversally converged, but with services and applications totally diverged next generationnetworks.Mehmet Erdas page 1
  • 2. Introduction:Network Planning Problem means in the broadest sense how to meet the customer,business and infrastructure specific conflicting competitive objectives under efficientresource-and capacity utilization constraints over time.The term „network planning“ has a broad scope of coverage, implying fundamentally both thestrategical -and the operational network planning processes, which can be summarized asthe radio network planning process tuned to the fixed-and mobile network planningprocesses , short-medium-and long-term resource allocation problem , capacity assignmentand routing problem together with the integrated network data base planning,considering thenetwork evolution, network compatibility and the integration of OEM-products into existingnetworks, network planning standardization, security and diffserv capability of a variety ofmiddleware applications developed independently of underlying network structures, but as afeedback influence increasing the complexity of network traffic load, assuming a unifiedfuture network described by ist network database. The treatment of telecommunicationsnetwork economics against the network availibility,network redundancy (availibility ofredundant paths meeting overload and burstiness in peak hours ), network security andnetwork database back-up and recovery concepts should be the main emphasis of anintegrated network planning. Shortly defined, the integrated network planning process coversthe orderly , time-dependent efficient deployment and management of computer-and datacommunications facilities. The new service product Network Planning is used in both theoperational ( referring to existing telecommunication networks) and the strategicalsense( identifying future technology trends driven by market forces ).The evolutionarynetwork planning aims at the overall technological-economical-and financial integration ofnew network components features and technologies into existing networks.For the development of network planning and integration tools, the well-known classicalO.R. algorithms of advanced.dynamic integer.programming and graph theory, queueingtheory and combinatorial optimization, branch and bound methods, penalty methods, andDiscrete Event Simulation.are the most commonly used network planning and optimizationalgorithms..(1),(2),(3),.(14)The Network planning problem has to be understood as an optimization problem, statedunder the optimization criteria of cost effectiveness, high reliability- availibility-flexibility,extendability of networks and network components to minimize the overall network costs(reduce equipment) with a modular subnetworks structure. The general planning factors tobe considered are mainly the technological factors, economical factors, financial factors,business factors, organizational factors,and environmental factors. Depending upon the levelof planning detail required , for the definition of the network components, network modules(HW-and SW), various system and network architectures, network services, networktopology ,different routing strategies specified according to the underlying network topologycan be used. Finally based on the assumed traffic load sharing and traffic channelizingmode, quite different problem formulations of network planning problem can be presented.The most important network features can be counted as the statistical multiplexing of loads,the existence of a large number of heterogenous subnetworks, their modular interconnection,and value-added services, the future growth prospects of existing networks The end –to –end delay, cell or packet loss, blocking probabilities and the calculation of the link budgetsand protocol overheads which are to be taken into account together with the quality ofservice parameters. The Bandwidth availibility on demand is another bottleneck, that has tobe considered in the formulation of network planning-and optimization problem.Theeconomies of scale, finiteness of resources, standardisation and growth prospects of newtechnologies, modular extendability, hardware and software variety and emergence of newsolutions should also considered as objective variables or constraints in the formulation ofheterogenous networks` planning problem.Mehmet Erdas page 2
  • 3. The evolution of networks over time is a key aspect of network planning and networkoptimization.The new network engineering requests may come due to a new product ortechnology or new customer (market-driven or technology-driven) expectations. Thedefinition of new services, business priorities, reuse of existing infrastructure in migration areimportant to identify the network design strategy.The optimized routing, colocation of network elements are the other factors influencing thecost-revenue-profit picture.Aggregation and/or Decomposition into smaller Problems:Methodologically,the network planning and optimization problem has to be divided into anumber of smaller, easily manageable subproblems. One set of subproblems might bedefined relying on the existing network structure, network topology, the priority rules forservices in proportion to their shares in revenues, and the routing strategy. Another set ofsubproblems might be defined as end-user terminal equipment design (intelligence of end-user equipment), access technologies design, the assessment of switching technologies andfinally the transmission systems.(5) The Markov Chains, queueing theory, general birth-deathprocesses and renewal theory can be used to unify the totally different world of circuitswitching and packet switching. The Erlang-k distribution and priority queueing models as M/G/m queues can be used efficiently to simulate the traffic load as Poisson distributedinterarrival times and service times; defining their ratio as the utilization factor. (6),(8).(9)Applications such as video and voice telephony are delay sensitive and will requiredifferentiated services (QoS) with prioritization introduced into the queueing models.Analysisof the accuracy of bursty traffic models together with the response and recovery times andload sharing/load balancing in case of overload are an essential part of performance analysisof telecommunication networks Traffic models should match closely to real data in order toobtain reasonable tracking of the critical network performance bottlenecks..Generically, one could optimize the cost-revenue-profit triple by minimizing the cost ofexpenditures for equipment and operations, and maximize revenues by introducing value-added intelligent services through intelligent networking (add a separate control layer toachieve service,network and end-user equipment independence) and doing all this over timeas technology, user requirements and the economic factors change. The decomposition ofnetwork planning problem into smaller optimization problems has to be done for the sake ofsimplicity, consistence, uniqueness and solvability.There are various types of classification approaches for different types of network planning,such as fixed and radio network planning, administrative planning, fundamental technicalplanning to develop plans for network management, switching and routing, addressing,signalling, operations, provisioning and maintenance. Engineering plans are detailed andimmediate plans. Another type of planning can be accomplished on the basis of networkcomponents selection, like the number of base stations, local exchanges, toll exchanges,interexchange transmission, loop plant, signalling network and customer premisesequipment, LAN, WAN,MAN, Routers, Bridges, Gateways etc. For GSM/GPRS/EDGE/UMTSnetwork planning, the main classification is usually the radio network planning and fixednetwork planning besides of course the packet switching and the circuit switching. Accordingto different services, another classification could be made as POTS, ISDN, SMDS or FRservices, Packet, Video, Cellular Telephone, E-Mail, Remote Login, File Transfer, ImageTransfer, Voice Connections, World Wide Web.According to timing or time coverage of plans, the long-term plans(5-20 years), medium-termplans (2-5 years), and short-term plans (1-2 years) could be done using iterative dynamicprogramming or simulation scenario techniques by changing the planning assumptions. ( 7)Mehmet Erdas page 3
  • 4. Performance Evaluation of High Speed Packet Switching NetworksThe Packet-switching network was developed during the 1960s.The idea behind a packet-switching network was to create a network of dedicated leased lines whose sole functionwould be to transport digital data traffic. At the source, data would be divided into groups ofbits called packets. An actual packet has two parts: Header and the actual information fieldor payload. The System Performance measures in a packet switched network are theinterarrival times, service times, queue length, transit time, waiting time, and server idle time.(2)The Header contains information about the originating point, packet`s destination, its priorityand its error codes. The payload is the group of information bits that has to be transportedover the network.algorithms running in the switching nodes read a packet`s destinationaddress and forward the packet over the next successive link on its way to its destination.The great advantage of statistical multiplexing in packet switching technology , that is sharingof transmission lines by the bursty data traffic between many users,lowered the cost oftransmitting data over leased lines and combined the inherent bursty data traffic intoaggregate flows that could be accommodated economically by long lasting leased-lineconnections. Today packet-switching is used overall in general user networks such asInternet as well as in specialized applications such as in establishing the connections intelephone networks through the Common Channel Signalling System 7.With the introduction of ATM (Asynchronous Transfer Mode) technology, the share of packet-switching in the total world communications bandwith increased drastically. ATM combinedbroadband(high-speed) communications and services of voice- data-and video traffic in anintegrated manner (ISDN). Some important advantages of ATM technology againstSTM(Synchronous Transfer Mode): No rigidly structured hierarchy anymore needed No time slot assignment (Mapping) problem anymore No need for separate switches at each data rate by multirate switching as a combination of 64kbps switching building blocks.Bursty data traffic and services instead of fixed-demand services possibleDuring Network Planning the individual network components are to be planned andintegrated into the existing GSM network and Internet. This covers the interconnecting ofnetwork equipment according to the network planning, configuration of system parametersfor each network component, customer specific setup of the network management systemtests with real applications and real traffic simulations.The scope of overall end-to-end network planning problem should be divided into smallersubnetwork planning problems as the-Radio Network Planning(RNP)-Fixed network Planning ( PSTN,B- ISDN..)-Mobile Network Planning (GSM/GPRS/EDGE/UMTS)-Database Planning (Backup-and Recovery (14)Planning a High Quality, High Performance Network ArchitectureThe right network architecture should be tailored depending upon the relative market choicesof companies; even within a single market the architecture and technology are to beconsidered as moving targets, under which we should look for optimum networksolutions(maximizing benefits while minimizing risks). Mobile operators are building networksonly for their own use, without any real traffic simulations. The large variety of subnetworksand services necessitates a dedicated and specialized planning. The diversity of hardwareand software complicates network management and planning. Therefore the choice of HWMehmet Erdas page 4
  • 5. and SW and the rapid growth in networks makes it compulsory to install higher capacitysystems accompanied by proper network planning.Mathematical Programming for Network PlanningAn objective function and associated set of constraints is called a mathematical program,consisting of decision variables and surplus or slack variables to convert the constraintinequalities into equations which are then to be solved by matrix operations of inversion andmultiplication. The set of all constraints determines the feasible solution space. The Objectivefunction might be cost, performance or reliability metrics. Network planning problemformulated as a mathematical programming, might have a single unique globally optimalsolution or many locally optimal solutions. A locally optimal solution is only optimal for alimited portion of the feasible solution space. Sometimes heuristic algorithms, which useintuitive procedures to find out optimal solutions, might be useful to achieve global optimalsolutions starting with local optimal solutions. The canonical problem formulation for networkplanning and queueing theory used to formulate the telecommunications network design andthe solution technique , called simplex algorithm, can be found in Ref.(1).p.14-41.The Network Optimization is indispensable because of shifts in subscriber-and applicationdistribution and their traffic behaviour, changes in the subscriber mobility profile, subscribergrowth, unbalanced market-driven regional network growth and limitations of frequencyresources on air-interface.Routing Problem and Discrete Event SimulationISPs or ASPs face a challenge in provisioning of network resources because of the rapidgrowth of bursty internet traffic and wide fluctuations of the traffic patterns. The dynamicrouting should be used to prevent congestions and application performance as a valuabletraffic engineering tool. The deployment of load-sensitive routing is however difficult due tooverheads imposed by link-state update propagation, path selection and signalling. Throughsimulation experiments of one week or one-month duration, packet flows could be traced todifferentiate between long-lived and short-lived flows to improve the performance of thelinks and to achieve the routing stability. The existing routing protocols OSPF, BGP, RIP etc.are optimizing in one way, leaving the longer paths underutilized.A middle approach between physical experimentation and statistical analysis which is oftenused, is simulation technique. Since simulations are performed with software , it is easy tochange or test the model assumptions, or change requests. The usual type of simulation of anetwork is called discrete event simulation. The”discrete events” are occurrences such aspackets being transmitted , a buffer receiving a packet, or a call being switched. Simulationscan be run to trace transient behaviour of networks,which occur over a very short period oftime as a result of some event. The behaviour of networks over long time periods and theself-similarity of internet traffic, that means the steady-state behaviour of networks could beobserved and simulated to examine various planning assumptions, whether they representthe reality.of traffic as it is. Discrete event simulation is stochastic in nature, because basicinputs like packet arrivals and call placements are to be generated randomly by usingpseudorandom number generators.as software products.(3)The self-similar traffic modelling is going to replace the poisson modelling of network traffic,because of long-range dependence in wide-area networks. The simplest models with long-range dependence are self-similar processes, which are characterized by hyperbolically –decaying autocorrelation functions. The long-range dependence of self-similar processes canbe charactereized by a single parameter, called the Hurst parameter., which can beestimated using Whittle`s procedure (11)Mehmet Erdas page 5
  • 6. Transient queueing analysis is essential for network planners to understand the temporalbehaviour of their networks. The sojourn time performance of a network node has to bestudied under realistic traffic environment. For that purpose , a network node has to bemodeled as a finite quasi-birth-death process(instead of simplest M/M/1queueing model) withlevel dependent transitions, which are used to model a controlled or prioritized queueingsystem, where both the arrival and the service processes are to be regulated based on theinstantenous buffer occupancy level, because the size of the buffer is always finite inreality.and the arriving cells are lost when the buffer is full. This approach allows theincorporation of more sophisticated and accurate traffic models than the previous 2/3 StateMarkov Models.of network traffic. The impact of input traffic characteristics and the effect ofvarious simplifying assumptions like infinite buffer approximations, the effect of statisticalmultiplexing and the controlling effect of preemptive cell discarding (to assure the QoS) onthe sojourn time behaviour of the system has to be studied further in depth.to explain thenodal congestion in networks planning. Realistic networks of today have large buffer size, butcomplex and bursty input traffic makes the infinite buffer assumption invalid. Bufferingtogether with the statistical multiplexing can be used to increase the redundancy, reliabilityand availibility of networks to avoid congestions and to provide the QoS parameters in caseof overload or highly bursty traffic input with long duration.(12, 13)Trends in Network Planning:For Transmission capacity services: TDM SONET/SDH WDM/DWDM-First step to the future–optical switching at 10-100 Terabits/sec.For Access Networks services: TDM CATV,DSL,802.11,LMDS Wireless-Mobile-IPConvergence of fixed-and mob.IP,100Mbit Ethernet is the right next step,but fiber optics isthe future transmission medium in telecommunications.For Fixed Voice Networks services: CS VoIP using H.323 Replace with SIP and MGCPSession Initiation and Mediagateway ProtocolFor Mobile Voice and Data Networks :.VoGSM SMS WAP GPRS EDGE UMTS or 802.11 IP-based new Value-Added Services, like IP-based Intelligent networking and new middlewareapplications development just by separating the control plane and data plane.New IP-Services Best Effort DiffServQoS IPsec.for VPN Security WDM-Switching replacingATM ; as a moving target between assured delay and assured bandwith use MPLS for trafficengineering, just putting the bandwidth where the traffic is or putting the traffic wherebandwidth is.Overall Trends and Conclusions: Fiber is the only future proof foundation for all networkservices; SIP and MGCP will be the key to voice/data convergence; mobile phone operatorswill become wireless Internet access providers and last but not least: Internet is able toprovide QoS and Security without Layer2 VCs. With the realization of UMTS, the cellularnetworks of the future might well be dimensioned for the dominant type of traffic which isexpected to be mobile data, rather than voice. This would lead to network consolidation of IP,ATM and Frame Relay.through network consolidating layers enabling cost savings ininfrastructure.FMC: Fixed-Mobile Convergence:The heterogenous networks evolution and the mobility of Internet requires a unique OAMConcept for common billing, operation and maintenance of diversified network services.FMC can be realized by establishing a combined switching centre enabling the service andsupport of both the mobile and the fixed customers through the same exchange. This mightbe a hardware or software solution or a combination of both depending upon the existingnetwork infrastructure. Global access to personalized services are independent of accessmethodology, underlying network and delivery method. It should be mentioned that theaccessnetwork is not so expensive to build out and to upgrade. Convergence will first happen inenterprise networks when voice is moved from traditional voice VPN (PABX networks) todata-VPN and thereafter into long distance IP-based intelligent VPNs. In the medium-term,Mehmet Erdas page 6
  • 7. the emerging technologies and standards will facilitate service and network convergence toan IP based network with fixed and mobile access increasing complexity with ever growingdata throughput rates, bandwidth allocation and network configuration managementproblems. The optimization criteria for such a converged network can be counted as the end-to-end targeted quality of service levels, throughput rates, link capacity utilization, theminimized overall cost and delay levels with differentiated security allocations for differentapplications and the interoperability or compatibility of hardware and software units withoutposing any difficulties for combined implementation.Conclusion:Modified overall network optimization problem formulation for a unified network planningprocessThe overall planning problem for such a converged network could be be formulated as theminimization of end-to-end total line costs (call set up, volume-and time dependant chargingaccounting for the cross-product of total connection time and volume of data transferred end-to-end packet-and circuit switched connections), subject to a given traffic load sharingmodel, given the chosen coding schemes for radio network coverage, given the specifiednode locations, inter-node and intra-node peak circuit- switched -and packet- switched trafficload sharing mechanisms, adjusted or matched by general birth-date stochastic queueingmodels delivering the required minimum link budgets and buffer sizes for smoothing out theburstiness of packet data traffic, over the decision variables of underlying network topologyand routing policy, yielding the total channel or link capacities adapted by channel allocationchoices (channelized, unchannelized, fractional, setting DE for FR or CLP for ATM orlabeling for ATM LSR ) relying on the DiffServ or prioritized QoS-using static,virtual, dynamicrouting mechanisms, without leaving any longer paths underutilized, if congestion in theshortest paths occurs), satisfying the allowed overall access-switching-transmission end-userequipment delays, reliability-redundancy-and availibility constraints, all being discrete(non-continuous) and iterated .over time covering the network planning period. The outputs ofsuch a planning model will be measured or scaled in multiple functional HW-or SW units,bits -and seconds, which are to be converted into monetary units using a market-drivensales, qualified cost- and pricing strategy allowing for the investment protection of investors-suppliers-and operators triple defined as survival value chain, such that none of the market-players will be threatened in survival.This overall problem formulation for the optimization of combined radio-and fixed networkplanning process could be extended for incorporating the involved database planning,database security, back-up and recovery processes.References:1-Thomas G. Robertazzi:Planning Telecommunication Networks , 1999, IEEECommunications Society, Ch.1-2-3 pp 2-362-Susan L. Solomon Simulation of Waiting Line Systems, 1983 Prentice Hall, pp 11-163-Jerry Banks,John S.Carson II, Barry L.Nelson :Discrete-Event System Simulation, 1999Prentice –Hallpp 92-96Mehmet Erdas page 7
  • 8. 4- J. Ioannidis, D. Duchamp and G.Q. Maguire Jr.: IP-based Protocols for MobileInternetworking.In Proc. SIGCOMM 91, ACM, Zurich, Sept. 1991, pp. 235-245.5- Pflug,G Stochastische Modelle in der Informatik, Stuttgart, 1986, p.85 and p.1176-Daigle, J..N.: Queueing Theory for Telecommunications, Addison-Wesley, 1992 pp- 6-13,Ch.3-47-Gupta V.P., „What is Network Planning“ IEEE Communications Magazine, Vol.23, Nr.10,Oct. 1985, pp 10-168-Kleinrock. L. Queueing Systems Vol 1-2, New York , 19759-Kleinrock, L; Queueing Systems-Problems and Solutions-, New York, 198910- Heinanen, J.“Futureproof network planning strategies“ International Conference inLondon, 24-25 May, 2000, organized by Vision in Business.11-Garrett.M and Willinger.W, : Analysis, Modelling and Generation of Self-Similar VBRVideo Traffic, in: Proceedings of SIGCOMM`94 , pp. 269-280, 199412- Kobayashi, H., Ren Q.: Nonstationary behaviour of statistical multiplexing for multipletypes of traffic, in : Proceedings of the 26th Annual Conference on Information Sciences andSystems, Princeton University Press, Princeton NJ, March 199213-Kant K., Introduction to Computer System Performance Evaluation, McGraw-Hill, NewYork, 1992.14-Kumar V.,Hsu M. Recovery Mechanisms in Database Systems, Prentice-Hall, Newjersey, 1998,pp. 56-68, 259-291, 661-697Mehmet Erdas page 8

×