Read more
This book answers a question which came about while the author was work ing on his diploma thesis [1]: would it be better to ask for the available band width instead of probing the network (like TCP does)? The diploma thesis was concerned with long-distance musical interaction ("NetMusic"). This is a very peculiar application: only a small amount of bandwidth may be necessary, but timely delivery and reduced loss are very important. Back then, these require ments led to a thorough investigation of existing telecommunication network mechanisms, but a satisfactory answer to the question could not be found. Simply put, the answer is "yes" - this work describes a mechanism which indeed enables an application to "ask for the available bandwidth". This obvi ously does not only concern online musical collaboration any longer. Among others, the mechanism yields the following advantages over existing alterna tives: - good throughput while maintaining close to zero loss and a small bottleneck queue length - usefulness for streaming media applications due to a very smooth rate - feasibility for satellite and wireless links - high scalability Additionally, a reusable framework for future applications that need to "ask the network" for certain performance data was developed.
List of contents
1. Introduction.- 1.1 Reader's Guide.- 1.2 Work done.- 2. Related Work.- 2.1 Congestion Control in Packet Networks.- 2.2 Problems and Issues.- 2.3 A Trend towards Explicit Signalling.- 2.4 Congestion Control in ATM: ABR.- 2.5 Quality of Service.- 2.6 QoS and Congestion Control.- 2.7 TCP-friendly QoS: A Comparative Evaluation.- 2.8 Explicit Traffic Signalling in Packet Networks.- 3. PTP: A Framework for Explicit Performance Signalling.- 3.1 End-to-end Arguments revisited.- 3.2 Architecture.- 3.3 Performance Parameters.- 3.4 Policy Considerations.- 3.5 The Performance Transparency Protocol.- 3.6 Content Type Definitions.- 3.7 Usage Scenarios.- 3.8 Realisation.- 4. Congestion Avoidance with Distributed Proportional Control.- 4.1 Architecture.- 4.2 Design Phase 1: Fluid Model.- 4.3 Analysis.- 4.4 Design Phase 2: Discrete Model.- 4.5 Evaluation 1: Dynamic Behaviour.- 4.6 Evaluation 2: Long-term Performance.- 4.7 Final Comments.- 5. Conclusion.- 5.1 Future Work.- Appendices.- A The ns Network Simulator.- A.1 Overview.- References.- About the Author.
Summary
This book answers a question which came about while the author was work ing on his diploma thesis [1]: would it be better to ask for the available band width instead of probing the network (like TCP does)? The diploma thesis was concerned with long-distance musical interaction ("NetMusic"). This is a very peculiar application: only a small amount of bandwidth may be necessary, but timely delivery and reduced loss are very important. Back then, these require ments led to a thorough investigation of existing telecommunication network mechanisms, but a satisfactory answer to the question could not be found. Simply put, the answer is "yes" - this work describes a mechanism which indeed enables an application to "ask for the available bandwidth". This obvi ously does not only concern online musical collaboration any longer. Among others, the mechanism yields the following advantages over existing alterna tives: • good throughput while maintaining close to zero loss and a small bottleneck queue length • usefulness for streaming media applications due to a very smooth rate • feasibility for satellite and wireless links • high scalability Additionally, a reusable framework for future applications that need to "ask the network" for certain performance data was developed.
