performance comparison of scheduling algorithms for iptv
play

Performance comparison of scheduling algorithms for IPTV traffic - PowerPoint PPT Presentation

Performance comparison of scheduling algorithms for IPTV traffic over Polymorphous OBS routers Jos e Alberto Hern andez V ctor L opez, Jos e Luis Garc a Dorado and Javier Aracil Networking Research Group Universidad Aut


  1. Performance comparison of scheduling algorithms for IPTV traffic over Polymorphous OBS routers Jos´ e Alberto Hern´ andez V´ ıctor L´ opez, Jos´ e Luis Garc´ ıa Dorado and Javier Aracil Networking Research Group Universidad Aut´ onoma de Madrid, Spain email: [Jose.Hernandez]@uam.es ICTON-MW’07 December, 2007 J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  2. Contents ◮ Introduction to OBS and to POBS ◮ IPTV traffic and its characteristics ◮ Scheduling algorithms for combining best-effort traffic with synchronous reservations for IPTV traffic: FF, RR, SRR, SSRR ◮ Scenario definition and experiments ◮ Conclusions J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  3. Optical Burst Switching review In OBS networks: ◮ Edge nodes aggregate incoming traffic into optical data bursts (1 optical burst contains many packets). ◮ Core nodes switch data bursts all-optically (no O/E/O conversion) accross the DWDM physical layer. J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  4. Burst-assembly process In an edge OBS node: ◮ For each data burst, a Burst-Control Packet (BCP) is generated and transmitted after the data burst is assembled. ◮ Typically, the BCP contains the size of the data burst and expected arriving time at the intermediate nodes. It reserves resources for the forthcoming data burst. ◮ The time difference between the BCP and its associated data burst is known as offset time . J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  5. Polymorphous OBS ◮ In the Polymorphous OBS architecture, a BCP (with extended attributes) may reserve: ◮ A full wavelength (1) ◮ Synchronous fixed-size time slots (2) ◮ Asynchronous best-effort capacity (3) Control Unit Control Plane BCPs 1 2 2 2 3 3 �� 1 T 2 2 2 3 3 �� 2 1 3 3 3 3 �� 3 Full wavelength reservation TDM circuit reservation OXC ◮ The synchronous reservations may be used by services which require periodic capacity, but do not need a full wavelength (for instance, IP TeleVision). Goal Find the best way to combine IPTV reservations with asynchronous best-effort traffic that brings best performance (less blocking probability). J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  6. IPTV Traffic characteristics ◮ Observed characteristics of IPTV service from a Spanish IPTV service provider: ◮ MPEG-2 encoding ◮ 4.16 Mbps per TV channel ◮ Constant Bit Rate stream with packet interarrival times of 2 , 5 ms ◮ This configuration belongs to Standard Definition TV J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  7. Scheduling algorithms for combining IPTV traffic together with best-effort ◮ First fit (FF): All TV channels are inserted together over the same wavelength. ◮ Round Robin (RR): TV channels are grouped together and allocated over different wavelengths following a Round Robin fashion. T on = N ch x T 1ch T on = N ch x T 1ch D D D D D �� 1 �� 1 T on +T off = 2.5ms T on +T off = 2.5ms D D D D �� 2 �� 2 ... ... D D D D D D �� M �� M First Fit Round Robin J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  8. Scheduling algorithms for IPTV traffic together with best-effort (2) ◮ Synchronous Round Robin (SRR): Channels are splitted into different wavelengths, and transmitted at the same time in each lambda. ◮ Synchronous Shifted Round Robin (SSRR): Same as SRR but the time is shifted for different lambdas. T on +T off = 2.5ms T on +T off = 2.5ms D D D D D D D ... ... �� 1 �� 1 T on = N ch x T 1ch / M D D �� 2 D D D D D ... ... ... T on = N ch x T 1ch / M �� 2 ... D D D D D D ... ... �� M �� M Synchronous Round Robin Synchronous Shifted Round Robin J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  9. Analytical approximations ◮ Using the Erlang-B formula, ignoring the retroblocking effect of OBS (which arises when offset times are highly variable): P block = P block | off P ( off ) + P block | on P ( on ) ◮ First Fit: T on + D E B ( I , M − 1) + T off − D P block = E B ( I , M ) T off + T on T off + T on ◮ Synchronous Round Robin: T on + D + T off − D P block = E B ( I , M ) T off + T on T off + T on ◮ D = burst size, M = Number of wavelengths, I = Offered traffic J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  10. Experiments: Scenario definition ◮ We have considered 192 channels transmitted over the POBS network. 192 × 4 , 16 Mbps = 799 Mbps over a M = 8-wavelength with capacity 10 Gbps per wavelength. ◮ T on + T off = 2 , 5 ms as measured, T off = 2 , 3 ms for FF and RR, and T off = 2 , 475 ms for SRR and SSRR. ◮ T on = N ch × B ch = 192 × 4 , 16 Mbps = 8 % × 2 , 5 ms FF, RR C 10 Gbps ◮ T on = N ch × B ch = 192 × 4 , 16 Mbps = 1 % × 2 , 5 ms SRR, SSRR M × C 8 × 10 Gbps ◮ Load: ρ = λ D M , where D ∈ { 0 , 025 , 0 , 125 , 0 , 25 , 0 , 5 ms } For instance, D = 150 packets × 1024 · 8 = 0 , 125 ms 10 Gbps J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  11. Simulations and results (1) D=0.025 ms D=0.125 ms 1 1 0.1 0.1 Blocking Probability Blocking Probability 0.01 0.01 0.001 0.001 1e-04 FF 1e-04 FF RR RR 1e-05 1e-05 SRR SRR SSRR SSRR 1e-06 1e-06 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 ρ ρ D=0.25 ms D=0.5 ms 1 1 0.1 0.1 Blocking Probability Blocking Probability 0.01 0.01 0.001 0.001 1e-04 1e-04 FF FF RR RR 1e-05 SRR 1e-05 SRR SSRR SSRR 1e-06 1e-06 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 ρ ρ ◮ FF and RR shows best performance, especially for large D . ◮ Hence, it is better to merge TV channel transmition than split them over different wavelengths. J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  12. Simulations and results (2) ρ =0.5 1 FF RR SRR SSRR Blocking Probability 0.1 0.01 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 D (ms) ◮ FF and RR shows best performance, and same behaviour regardless of burst size D . J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  13. Simulations and results (3) D=0.025 ms D=0.125 ms 1 1 p s p s p t p t 0.1 0.1 Blocking Probability Blocking Probability 0.01 0.01 0.001 0.001 1e-04 1e-04 1e-05 1e-05 1e-06 1e-06 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 ρ ρ D=0.25 ms D=0.5 ms 1 1 p s p s p t p t 0.1 0.1 Blocking Probability Blocking Probability 0.01 0.01 0.001 0.001 1e-04 1e-04 1e-05 1e-05 1e-06 1e-06 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 ρ ρ J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  14. Simulations and results (4) D=0.025 ms D=0.125 ms 1 1 p s p s p t p t Blocking Probability Blocking Probability 0.1 0.1 0.01 0.01 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 ρ ρ D=0.25 ms D=0.5 ms 1 1 p s p s p t p t Blocking Probability Blocking Probability 0.1 0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 ρ ρ J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  15. Summary and conclusions ◮ Main conclusions: ◮ FF and RR show similar performance results and are the best strategies among the four scheduling algorithms studied. ◮ It is better to group all channels and transmit them at once than splitting them over different wavelengths, especially when D is comparable to T off . ◮ The analytical approximations for FF and SRR are shown to be accurate. J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

  16. Questions Thank you!! J. A. Hern´ andez et al Scheduling algorithms for IPTV in POBS

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend