Chapter5 Traffic Engineering Topics Background Throughput calculations Packets

Chapter-5 Traffic Engineering

Topics § Background (Throughput calculations). § Packets, frames and cells per second. § Traffic Engineering Basics: Traffic characteristics and source models. § Statistical Behavior of User Traffic § Poisson Arrivals and Markov Processes: § Statistical Behavior of User Traffic, Voice traffic Modeling § Queued data and packet switched traffic modeling § Causes of Delay , Availability , Reliability and Maintainability Network Design Tools § Physical and Statistical Design

Background (Throughput calculations) Before going to the topic of traffic engineering, let us know about throughput. Throughput is the actual amount of user data that is transmitted over the access circuit [Internet] and received by the network node [user]. Every user access circuit is regulated by throughput. Throughput can also be measured end to end across the entire network. The common units of measurement for throughput are bits, bytes, packets, frames and cells per second.

Packets, frames and cells per second • The term packets, frames and cells per second are used throughput the calculation , but to simplify the discussion. • The maximum achievable packet per second (PPS) rate on physical circuit can be calculated as follows:

Traffic Engineering Basics: Traffic characteristics and source models • Source Model Traffic Parameter Characteristics – There are two basic philosophies for characterizing source traffic parameters : Deterministic and probabilistic. • Deterministic parameters are based upon a specific traffic contract, with conformance verifiable on a unit by unit basis. – The probabilistic (also called stochastic) model is typically measurable only over a very long-term average. • Since the method and interval for computing the average can differ, conformance testing defines the details of the measurement method. • Specification of the statistical model is also required.

Traffic Engineering Basics: Traffic characteristics and source models • General Source Model Parameters: Burstiness is a commonly used measure of how infrequently a source send traffic. Source that infrequently sends traffic is said to be very bursty. The formula for Burstiness –

Traffic Engineering Basics: Traffic characteristics and source models • Poisson Arrivals and Markov Processes: Random arrival processes are described in general and the Poisson (Or Markov) process in particular with reference the following figure.

Traffic Engineering Basics: Traffic characteristics and source models • Poisson Arrivals and Markov Processes…. . If t : interval of time and : arrival rate per second This is called a memory less process, because the probability that the interval of time will be x second is independent of memory of how much time has already expired.

Statistical Behavior of User Traffic • Traffic pattern can be calculated in many ways. User information arrives at the network node based on the statistical arrival rates. • Typically , not all users need to communicate with all other user at the exact same time, therefore statistical approximation can be used to model these traffic patterns.

Voice traffic Modeling (Erlang Analysis) The standard for statistically calculating an estimate for user demand based on random call arrival and holding was modeled by A. K. Erlang. The unit of measure called the erlang is calculated as: Where = the call arrival rate in calls/hour and = average holding time in hours

Voice Traffic Modeling Voice Network

Queued data and packet switched traffic modeling • While Erlangs work well predicting voice network and circuit-switched traffic rates, they do not work well with packet-switched networks. – In packet-switched networks, some level of queuing is employed so that packets are queued in buffers and transmitted when congestion ceases, rather than being immediately blocked. – Packet-switched networks provide a mix of protocol and traffic types, whereas voice and circuit-switched networks provide point-to-point, transparent homogeneous transport of information. • Therefore, packet switching demands a different analysis of traffic handling

Designing for Peak When calculating the arrival rate of calls , a figure was called busy hour. Any network must be designed to handle the busiest traffic periods, or data will be blocked or lost

Causes of Delay Causes of delay in network: §Propagation path delay §Line speed §Number of hops between routing or switching nodes §Hardware and Software interface buffers §Load on every component across the path §Windows size §Memory and buffers §Pad functions §Address or route database §Security §Changes in traffic load §Filtering , forwarding and processing packets frames and cells

Availability and Reliability • Two quality measures of hardware and software must be known by the network designer that is availability and reliability. • These two measures are depend on MTBF (Mean Time Between Failure) and MTTR (Mean Time To Repair) this is given by the vendor.

Availability • Availability: Availability is the amount of time the system is working when compared to the measured lifetime of the system. • Availability is calculated by:

Reliability

Maintainability • Maintainability: Maintainability is defined as the measure of effort a user or network provider must put into the network to maintain its operating efficiency. • Maintainability includes. – Defining how the designer will expand as requirements increase – Expandability and flexibility of the hardware platform – Ease of software upgrades • Build in tools for network monitoring, management, troubleshooting and reporting.

Designing Tools • Cost model: calculate the cost of the transport , hardware, and software portions of the network. • What if analysis: Recommend design changes based on failure or network configuration changes • Failure analysis: Describes the completely the effects on demand traffic of failure of some portion of the transport network or devices within the network

Designing Tools… • Visualize: Show the Network Elements and connection on the screen and on paper. • Simulate: Models the operation of the network devices with respect physical, data link, network layers. • Characterize: calculate key performance (KPIs) that describes design, efficiency utilization and so on

Designing Tools… • Optimize: changes the network design to achieve the higher utilization or performance and models cost trade-offs. • Design: produces transform demand from profiles • Report: produce utilization, link , equipment , site reports.

Model for Physical Design

Model for Statistical Design