Modeling the performance of DCF in mesh networks Andrew Symington, DNA Group
Data Network Architectures Group2 Presentation Outline Background Wireless Networks and IEEE The Distributed Coordination Function Performance Analysis & Modeling Modeling the Losses Incurred by DCF The IEEE Test Bed The Bigger Picture Questions
Data Network Architectures Group3 My Background & Future Business Science (Computer Science) at UCT Currently pursuing an MSc in DNA Group Would like to continue to PhD in the UK No long-term academic path (yet) Plan for the future –Embedded Systems –Wireless Telecommunication –Or a combination of the two!
Data Network Architectures Group4 DNA Group Research Areas Group is fairly new (<2yrs) to wireless networks Wireless research focus currently on : –General Mesh Networks and wireless Internet delivery – Security OLSR - previous talk – Performance Modeling DCF
Data Network Architectures Group5 Wireless Topologies Possible Topologies Point-to-point (line replacements) Point-to-multipoint (infrastructure) Multipoint-to-multipoint (ad hoc) –Mobile ad hoc networks (MANETs) –Mesh networks Infrastructure Ad hoc
Data Network Architectures Group6 IEEE Background Set of standards guiding WLAN development Split OSI Data Link layer into MAC and LLC IEEE n in draft - promises 248 Mbps –Theoretical vs. Actual rates differ LLC Flow control & Multiplexing MAC MAC-level Security Distributed Coordination Function Point Coordinator Function Enhanced Distributed Channel Access PHY IR FHSS DSSS a OFDM b HR-DSSS g OFDM
Data Network Architectures Group7 The Distributed Coordination Function Medium Access Control technique Ensures equal, but NOT fair, access to the channel A form of CSMA/CA –CSMA/CD not possible with wireless –Uses Exponential Binary Back-off Inter-frame Spaces define frame priority Designed for a Best-Effort service –No service guarantee –No QoS for multimedia
Data Network Architectures Group8 Factors Affecting DCF Collisions Recovery –Hidden Nodes –Exposed Nodes RTS / CTS –Longer Handshake –Reduces Error Increased Nodes Increased Traffic Topology
Data Network Architectures Group9 My Research Focus “ Modeling the performance of the Distributed Coordination Function in mesh networks” –Performance : Degree of QoS provided –Acknowledge that DCF is inefficient Thus, existence of e –Exacerbated by mesh networks Number of hidden / exposed nodes –But, by how much? And can we predict it? –Pave the way for EDCA analysis
Data Network Architectures Group10 What is Quality of Service? “A set of qualities relating to the collective behaviour of one or more objects” - ITU X.605 More specifically, for IEEE –Throughput –Response time –Jitter –Packet loss Increasingly NB for multimedia IEEE e and IEEE T (draft) –Not widely-adopted
Data Network Architectures Group11 Modeling Process Developing an analytic model is not a trivial process It’s NOT Simulation! Workloads NB –Trace –Synthetic
Data Network Architectures Group12 Existing Analytic Models for DCF Bianchi’s Model is widely-adopted DCF back-off modeled as Markov process Assumes a constant collision probability p Stage Back-off Counter
Data Network Architectures Group13 Bianchi’s Model p is a function of the number of contending nodes Using the Markov model, Bianchi derives : –The probability of successfully transmitting in a randomly chosen slot time –And, using , derives the normalised expected throughput at a node Normalised Throughput –Fraction of the base PHY rate offered –Indicates losses due to DCF overhead
Data Network Architectures Group14 Integration with MicroSnap Tool written as a DNA Project MicroSnap models stochastic queuing networks Environment Comprised of –Service Centres –Workloads –Traffic Classes –Routes (static) Program in MicroSnapL interface language
Data Network Architectures Group15 Fitting Models to MicroSnap Convert wireless links service centers
Data Network Architectures Group16 Workflow Diagram CONCLUDE DERIVE MACHINE MODEL FOR MESH DCFGENERATE TEST CASES (WORKLOADS/METRICS) EXPERIMENTAL (TEST BED) COMPARE RESULTS ANALYTIC (MICROSNAPL) SIMULATION (OMNET++)
Data Network Architectures Group17 The Mesh Test Bed The DNA Group is currently assembling an multi-hop wireless mesh network The DNA Group feels that the test bed will –Compliment research –Strengthen results –Provide a means to implement and test concepts –Attract students to the field
Data Network Architectures Group18 Progress on the Test Bed Short term goal of 9 nodes (double within a year) Already purchased a large portion of the hardware Challenges –PCI 2.1 / 2.2 Incompatibility –Limiting Signal Noise Injectors Attenuators –Interference More questions? Speak to me afterwards!
Data Network Architectures Group19 The Bigger Picture DNA Group is developing a suite of tools for wireless network performance modeling –Analytical –Simulation –Experimental (via test bed) Paolo Pileggi currently developing the framework as part of Honours project The work done in my research will contribute to a module within the IEEE MAC More specifically, it will assist in generating the machine model for arbitrary mesh networks
Data Network Architectures Group20 Thank you! Questions?