IntroductionWhy is it important?PowerPoint PresentationAnswerExampleA typical scenarioTaskOther issuesRequirements in practice (example)Slide 10Slide 11Slide 12Slide 13Slide 14Network Planning ProcessSlide 16Slide 17Slide 18DimensioningLoad BalancingSlide 21Network MonitoringWhy is network planning hard?Ways outDecomposition ExampleSample problem (arises as part of topology design)More questionsEven moreHow about this network?Analysis vs. Design (Synthesis)A sample topology design problemWhat if even a subproblem is too hard? (it can easily happen, as the examples suggest)Introduction•Main subject of course: methods of planning (design) of communication/computer networks•What is network planning?–In one sentence: mapping given requirements into a network.•In other words: find the network design that satisfies the requirements with as low cost as possible.1Why is it important?•A network (especially a large one) is a big investment.•Even few percents of cost savings can make a significant difference.•The quality of the network for a company or institution may have a strong influence on how well the business can go.•The design may also have a strong effect on future investments (extension, upgrade, changing requirements etc).2ExampleAnswerExample•The above example is for illustrative purposes.•A similar argument holds for optical cables–Refer to the bust of early 2000s … too much optical fiber5A typical scenario•Given requirements/input: •Traffic demands among various sites (traffic matrix)•Traffic characteristics (constant or variable rate, delay tolerance, burstiness etc)•Performance requirements (delay and loss bounds, data rate guarantees, etc)•Conditions on network topology (example: should provide backup route in case of a link/node failure)•Cost structure•Other: reliability, security, manageability, business and political priorities etc.•Available information is often incomplete, inaccurate, indirect.•Finding full information may be infeasible•Often needs to rely on (rough) estimations.6Task•Given the requirements (known or estimated), find: –Network topology: what to connect to what–Technology: what kind of equipment and links to use –Dimensions (capacities of links and equipment)–Map traffic onto network (flow routing)–Implementation plan (from which vendors to buy equipment, Internet service etc.)7Other issues•“Future proof” design: plan the network to serve the current demand p lus the projected future demand increment within the planning time horizon.–Again the telecom bust of 2000s.–Dilemma: •Short term view: minimize cost for now. May need costly upgrade soon.•Long term view: “overdesign” now with extra cost, but save more later, on the long term.•Security•Reliability•Manageability•Business and political priorities 8Requirements in practice (example)9Initial requirements10How do we derive actual traffic demands from this, if the management does not provide more info?…We may feel we never solve the design problem…Answer: Never say never!11User questionnaire12Second iterationof requirements13Requirements map14 Another example of a requirements mapNetwork Planning Process•Network planning process involves three main steps:•Topological design: This stage involves determining where to place the components and how to connect them. The (topological) optimisation methods that can be used in this stage come from an area of mathematics called Graph Theory. These methods involve determining the costs of transmission and the cost of switching, and thereby determining the optimum connection matrix and location of switches and concentrators.[1]15Network Planning Process•Network-synthesis: This stage involves determining the size of the components used, subject to performance criteria such as the Grade of Service (GOS). The method used is known as "Nonlinear Optimisation", and involves determining the topology, required GoS, cost of transmission, etc., and using this information to calculate a routing plan, and the size of the components.[1]16Network Planning Process•Network realization: This stage involves determining how to meet capacity requirements, and ensure reliability within the network. The method used is known as "Multicommodity Flow Optimisation", and involves determining all information relating to demand, costs and reliability, and then using this information to calculate an actual physical circuit plan.[1]17•These steps are interrelated and are therefore performed iteratively, and in parallel with one another. 18Dimensioning•The purpose of dimensioning a new network/service is to determine the minimum capacity requirements that will still allow the Teletraffic Grade of Service (GoS) requirements to be met. To do this, dimensioning involves planning for peak-hour traffic, i.e. that hour during the day during which traffic intensity is at its peak.•A dimensioning rule is that the planner must ensure that the traffic load should never approach a load of 100 percent19Load Balancing•Load balancing is a computer networking methodology to distribute workload across –multiple computers or a computer cluster,– network links, –central processing units, –disk drives, or other resources, to achieve optimal resource utilization, maximize throughput, minimize response time, and avoid overload. 20Load Balancing•Using multiple components with load balancing, instead of a single component, may increase reliability through redundancy. •The load balancing service is usually provided by dedicated software or hardware, such as a multilayer switch or a Domain Name System server.–Eg …21Network Monitoring•The term network monitoring describes the use of a system that constantly monitors a computer network for slow or failing components and •that notifies the network administrator (via email, sms, pager or other alarms) in case of outages. •It is a subset of the functions involved in network management22Why is network planning hard?•We want to find a design that minimizes the cost, such that all requirements are satisfied.•Impossible to try out all possible design choices –Typically there is an astronomical number of possibilities (practically infinite)–Efficient algorithms are needed, looking for clever shortcuts.•Generates very hard optimization problem–It is often hard to come up even with a good formulation!–Finding a globally optimal solution is
View Full Document