Skip to content

9.13.md

Latest commit

 

History

History
39 lines (23 loc) · 2.56 KB

9.13.md

File metadata and controls

39 lines (23 loc) · 2.56 KB

End-to-end Arguments in System Design

J.H. Saltzer, D.P. Reed and D.D. Clark

What is the Problem?

Since the lower level is unaware of the requirements of the higher-level application, such low-level implementations may be redundant or inefficient when considering the costs of providing them.

Summary

For the proposed "end-to-end arguments", the authors argue that higher-level functionalities should be implemented by the end hosts because they are familiar with the exact application requirements. On the other hand, the lower-level mechanisms on which these higher-level functionalities are based should be flexible and basic building blocks that only focus on performance enhancements. Finally, the authors discuss the pros and cons of their idea by analyzing detailed examples.

Key Insights

  • To move functionalities from lower level to higher level, the system designer should consider the tradeoff between the reliability and the performance carefully.
  • While applying the end-to-end argument, identifying the appropriate end points requires subtlety of analysis on the application requirements.

Notable Design Details/Strengths

  • The authors do not insist that the end-to-end arguments suit for all the scenarios. They suggest carefully reviewing the application requirements before applying their proposed design principle. Due to the performance reasons, placing some functionalities in the lower level may be beneficial.
  • The end-to-end arguments may have a guiding significance for the layered communication protocols.

Limitations/Weaknesses

  • The end-to-end arguments do not take other design criteria into consideration except the reliability and the performance, such as the communication security. Sometimes, it is necessary to include some basic security protections at the lower layer. Otherwise, the applications need to implement all the protection techniques on their own.
  • Although this paper focuses on the discussion about the design principle - the end-to-end argument, it does not provide any statistics to show its effectiveness.

Summary of Key Results

  • This paper proposes a design principle, named "end-to-end argument", providing guidance for the choice of function placement among different modules of a distributed computer system.
  • This paper also discusses detailed examples to analyze the pros and cons of the end-to-end arguments.

Open Questions

  • To explore the design or implementation details about the layered communication protocols.
  • It is worth considering a one-for-all solution for the communication system design.