html5-img
1 / 12

Performance Analysis of the IPACT Dynamic Bandwidth Allocation Algorithm for EPONs

Performance Analysis of the IPACT Dynamic Bandwidth Allocation Algorithm for EPONs. Joint research between IBBT-UGent and ENST-Paris. Passive Optical Networks. Downstream (P2MP): Packets filtered in Optical Networks Units (ONUs) Upstream (P2P): MAC protocol to avoid collisions

karl
Download Presentation

Performance Analysis of the IPACT Dynamic Bandwidth Allocation Algorithm for EPONs

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Performance Analysis of the IPACT Dynamic Bandwidth Allocation Algorithm for EPONs Joint research between IBBT-UGent and ENST-Paris Dept. of Information Technology (INTEC)

  2. Passive Optical Networks • Downstream (P2MP): • Packets filtered in Optical Networks Units (ONUs) • Upstream (P2P): • MAC protocol to avoid collisions • Offers an arbitration mechanism for upstream traffic • Provides signaling infrastructure • Bandwidth allocation algorithm to divide bandwidth Performance analysis of the IPACT DBA Algorithm for EPONs Dept. of Information Technology (INTEC)

  3. Performedresearch (IBBT & ENST) Furtherresearch Ethernet PON (EPON) • MAC protocol • EPON uses Multi Point Control Protocol (MPCP) • Ethernet control messages: GATE and REPORT • Dynamic band allocation (DBA) algorithms: 2 examples • IPACT (Interleaved Polling with Adaptive Cycle Time) • OLT keeps a polling table • Contains amount of bytes in each ONU and RTT • Updated by examining each ONU via round-robin • OLT grants a transmission window to each ONU • BGP (Bandwidth Guaranteed Polling) • OLT guarantees certain ONUs with a fixed bandwidth Performance analysis of the IPACT DBA Algorithm for EPONs Dept. of Information Technology (INTEC)

  4. Original model (ENST) • Analytical model of IPACT • ONU-architecture supporting three service classes • Expedited ForwardingAssured ForwardingBest Effort • Modeled by Weighted Round Robin (WRR) • Shortcomings ( model differs from reality) • Some oversimplifications in analytical IPACT model • WRR only suitable for a permanently available link • WRR not compliant with IEEE 802.1D standard Performance analysis of the IPACT DBA Algorithm for EPONs Dept. of Information Technology (INTEC)

  5. New model (IBBT, ENST) • Problem original model: • Mathematical description differs from reality • Model tries to model too much at once: • IPACT itself already very complex • Differentiated services even more complex • New Approach: • Analyzing simpler cases to provide real insight into the matter • Traffic load: symmetric (i.e. equal for each ONU) • Packet arrivals: Poisson distributed • Packet sizes: constant fixed length • No implementation of differentiated service classes • Validation verified by simulations Performance analysis of the IPACT DBA Algorithm for EPONs Dept. of Information Technology (INTEC)

  6. Analysis of IPACT algorithm • Analysis consists of two steps • Thorough analysis of the cycle times • Cf. IPACT, Interleaved Polling with Adaptive Cycle Times • Cycle time = time between the start of two successive transmission windows for a fixed ONU. • Derivation of the packet delay from the cycle times • Several service disciplines exist in IPACT • Service discipline = possibility for the OLT: • to determine the granted window size Wi • dependent on the requested window Vi • Three service disciplines analyzed: • Fixed: Wi = WMAX • Gated: Wi = Vi • Limited: Wi = min(Vi , WMAX) Performance analysis of the IPACT DBA Algorithm for EPONs Dept. of Information Technology (INTEC)

  7. Simulations in NS-2 • Simulation parameters • N Number of ONUs 16 • λ Load per ONU 5 → 65 Mbit/s (Poisson – symmetric) • ℓ Distance ONUs - OLT 20 km • TguardGuard time 5 or 1.5 µs • B Packet size 12000 bits • Breq REPORT message size 576 bits • Beth Ethernet overhead 304 bits • RU Upstream bandwidth 1 Gbit/s • PmaxMax. transmission window 10 or 3 packets Performance analysis of the IPACT DBA Algorithm for EPONs Dept. of Information Technology (INTEC)

  8. Fixed service • Constant cycle time Tcycle • Average queue size at discrete moments nTcycle • Queue sizes at t = nTcycle form a discrete homogeneous Markov chain  transition probabilities can be defined • Probability formulas: obtained by Poisson properties • Stationary distribution: obtained by linear system of equations • Delay derived from queue sizes (Little’s law) Performance analysis of the IPACT DBA Algorithm for EPONs Dept. of Information Technology (INTEC)

  9. Gated service • Varying cycle times: more complex analysis • Successive cycle times influence each other (correlated) • Distinction between low and high traffic load • Low: cycle time mostly determined by the traffic of that ONU (+ in case of clustering, the ONUs that are polled right before) • High: cycle time most often determined by the traffic off all ONUs together (distribution of cycle times via Markov chain) • Delay directly derived from average cycle time Performance analysis of the IPACT DBA Algorithm for EPONs Dept. of Information Technology (INTEC)

  10. Limited service • Some properties similar to fixed service and some similar to gated service • Possibility that a packet cannot be sent with its first requested window (cf. fixed) • Cycle time is variable (cf. gated) • System too complex for a complete analytical analysis similar to fixed or gated service • Especially in case of a small transmission window and a high load (high chance that a packet cannot be immediately sent) • In the other cases, a model analogue to gated service gives fairly good results Performance analysis of the IPACT DBA Algorithm for EPONs Dept. of Information Technology (INTEC)

  11. Extensions • Asymmetric load • Fixed: analysis still applies • Gated: analysis for high load is valid • Distribution packet size • Fixed: difficult since an unused part of the transmission window • Gated: good results obtained by using an average packet size • Self-similar traffic • too complex for analytical model Performance analysis of the IPACT DBA Algorithm for EPONs Dept. of Information Technology (INTEC)

  12. Conclusion • Analytical model formulated for IPACT • Fixed and gated service: valid model • Starting from the cycle time (very important for gated service) • With the assumptions of symmetric traffic load, Poisson traffic and fixed packet size • Limited service: too complex for numerical analysis • Also a valid model in some more general cases • The importance of simulations • Allows to verify analytical method and limitations • Only possibility in more complex situations • E.g. for the moment: research to the possibilities of the Bandwidth Guaranteed Polling DBA algorithm Performance analysis of the IPACT DBA Algorithm for EPONs Dept. of Information Technology (INTEC)

More Related