Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 Modeling and Performance Evaluation of DRED (Dynamic Random Early Detection) using Fluid-Flow Approximation Hideyuki Yamamoto, Hiroyuki Ohsaki Graduate.

Similar presentations


Presentation on theme: "1 Modeling and Performance Evaluation of DRED (Dynamic Random Early Detection) using Fluid-Flow Approximation Hideyuki Yamamoto, Hiroyuki Ohsaki Graduate."— Presentation transcript:

1 1 Modeling and Performance Evaluation of DRED (Dynamic Random Early Detection) using Fluid-Flow Approximation Hideyuki Yamamoto, Hiroyuki Ohsaki Graduate School of Information Sci. & Tech. Osaka University, Japan hideymmt@ist.osaka-u.ac.jp

2 2 1. State Goals and Define the System Goals Confirm validity of our approximate analysis Investigate DRED’s steady-state/transient-state performance System Definition IP network including... DRED routers and links source and destination hosts

3 3 2. List Services and Outcomes Services Provided Congestion control for TCP flows Outcomes High link bandwidth utilization? Low packet loss probability? Low packet transfer delay/jitter?

4 4 3. Select Metrics Speed (case of successful service case) Individual TCP throughput, round-trip time, packet loss probability Global Queue occupancy, link utilization, packet loss probability Reliability (case of error) None Availability (case of unavailability) None

5 5 4. List Parameters System parameters Network related Topology Link bandwidth, latency, loss ratio DRED router related Control parameters (  t  α, β, θ, T, L) Queue size Workload parameters # of TCP flows, TCP traffic pattern Background traffic pattern

6 6 5. Select Factors to Study System parameters Network related Topology Link bandwidth, latency, loss ratio DRED router related Control parameters (  t  α, β, θ, T, L) Queue size Workload parameters # of TCP flows, TCP traffic pattern Background traffic pattern

7 7 6. Select Evaluation Technique Use analytical modeling? Yes Use simulation? Yes Use measurement of real system? No

8 8 7. Select Workload TCP flows Persistent traffic # of TCP flows: 1 -- 1000 Background traffic 0 -- 70% of the bottleneck link bandwidth

9 9 8. Design Experiments

10 10 9. Analyze and Interpret Data

11 11 10. Present Results

12 12 Network Topology R1R2 :::: :::: Source 1 Source N Sink 1 Sink N c [packet/ms] 100Mbps 10ms 100Mbps 10ms source Sink Router dred DRED Queue TCP Flows Identical round trip time [ms]


Download ppt "1 Modeling and Performance Evaluation of DRED (Dynamic Random Early Detection) using Fluid-Flow Approximation Hideyuki Yamamoto, Hiroyuki Ohsaki Graduate."

Similar presentations


Ads by Google