Scalable QoS Provision Through Buffer Management

1. Scalable QoS Provision Through Buffer Management Roch Guerin Sanjay Kamat Vinod Peris Raju Rajan IBM Thomas J. Watson Research Labs

2. Outline • Differentiated Packet Treatment for QoS • Performance Objectives • Design Space Scheduling and Buffer Management Schemes • Comparing FIFO vs WFQ -- Worst case buffer tradeoffs • Examining tradeoffs with strict buffer partitioning • Examining tradeoffs with buffer sharing • Hybrid Schemes • Conclusions

3. Differentiated Packet Treatment Buffer Management Routing and Classification Scheduling Policing • Flows (Unit of service guarantee -- varying granularity) • QoS Resources -- Link capacity and Buffer Space • Scalability • Processing time per packet versus number of flows • State size versus number of flows Discard Discard

4. QoS Performance Objectives Throughput maximization . Flow Isolation & rate guarantees Fair allocation of excess resources Implementation cost Memory Processing

5. QoS Allocation Schemes No Buffer Management (FIFO) • Scheduling • FIFO • WFQ • Buffer Management • No Buffer Management (FIFO) • Buffer Partitioning • Buffer Sharing • Fair Excess Capacity distribution • (Choudhury and Hahne) . Buffer Partitioning headroom Buffer Sharing Empty buffer spaces Excess capacity sharing & fairness

6. Design Space . FIFO Buffer Management Fair and flexible excess capacity distribution Per-Flow Scheduling (WFQ) FIFO Scheduling Rate Based Per Flow Buffer Management FIFO treatment of excess traffic

7. Benchmarks WFQ scheduling & Buffer Management . FIFO scheduling & Buffer Management WFQ Scheduling & No Buffer Management FIFO scheduling & No Buffer Management

8. Rate Guarantees through buffer management alone Reservation Buffer Partition Offered Service Size Load Achieved r bits/sec Br/R bits Conformant Lossless Non-conformant Losses limited by non-conformance (s bits, r bits/sec) s+ Br/R bits Conformant Lossless Non-conformant Losses limited by non-conformance Partitioned Buffer Buffer Size B bits Link Capacity R bits/sec FIFO Scheduler

9. Buffer Partitioning w. FIFO vs. WFQ(Worst case comparison) R Sri WFQ Admissibility Checks Bandwidth availability B  Ssi Buffer availability R Sri Bandwidth availability B  Ssi+ (Sri ) B/RBuffer availability Equivalently B Ssi/(1-utilization) FIFO Admissibility Checks

10. Experimental Setup Link Scheduler On-off source Regulator Buffer Manager 500KB to 5MB 48Mbps Flow Token Token Mean Peak Number bucket Rate Rate Rate 0 50 kB 2Mbps 2Mbps 16Mbps 1 50 kB 2Mbps 2Mbps 16Mbps 2 50 kB 2Mbps 2Mbps 16Mbps 3 100 kB 8Mbps 8Mbps 40Mbps 4 100 kB 8Mbps 8Mbps 40Mbps 5 100 kB 8Mbps 8Mbps 40Mbps 6 50 kB 0.4Mbps 4Mbps 40Mbps 7 50 kB 0.4Mbps 4Mbps 40Mbps 8 50 kB 2Mbps 16Mbps 40Mbps

11. Benchmarks WFQ scheduling & Buffer Management . FIFO scheduling & Buffer Management WFQ Scheduling & No Buffer Management FIFO scheduling & No Buffer Management

12. Aggregate ThroughputBuffer Partitioning vs. No Buffer Management Throughput (Mbps Buffer Size (Mb)

13. Losses for conformant flowsBuffer Partitioning vs. No Buffer Management Loss (percent) Buffer Size (Mb)

14. Aggregate ThroughputBuffer Sharing vs. No Buffer Management Throughput (Mbps) Buffer Size (Mb)

15. Excess Capacity Sharing Throughput (Mbps) Buffer Size (Mb)

16. The Hybrid Scheme Flow -- Unit of buffer allocation Queue -- unit of bandwidth allocation WFQ Scheduler Protect individual flows while sharing buffers “Allocation” & “Headroom” Fair Excess Capacity distribution based on “holes”

17. ThroughputIntegrated Scheme Throughput(Mbps) Buffer Size (Mb)

18. Flow Isolation and Rate GuaranteesIntegrated Scheme Loss (percent) Buffer Size (Mb)

19. Excess Capacity DistributionIntegrated Scheme Throughput (Mbps) Buffer Size (Mb)

20. Conclusions • The Integrated Scheme • Scheduling on a fixed number of “Queues” and buffer management on the finer granularity of “Flows” • Rules of thumb for flow grouping • Buffer sharing & flow isolation within a single • use “headroom” to limit impact on conformant flows in buffer limited system • Empty buffers (holes) to regulate the sharing of excess bandwidth • flexible notion of fairness • Flexibility to choose tradeoffs depending on operational environment • Some quantitative guidelines on the choice of design and setting of parameters