Network Systems Congestion Control Study Notes.

Network Systems Congestion Control Notes

Introduction

  • Presentation by Pieter-Tjerk de Boer.
  • Some images copyright 1996-2012 J.F. Kurose and K.W. Ross, and/or copyright 2010 Elsevier Inc.
  • Course: UT/EWI - Network Systems 2026
  • Upcoming timeline:
    • Only two weeks until the 3rd exam.
    • Exam covers less material and is shorter (1.5 hours).
    • More time available as there's less math/programming for TCS/EE this week.
  • Topics of focus:
    • This week: resource allocation.
    • Next week: security.
  • Reminder: students tend to find this exam harder than the first two.

Resource Allocation Theme

  • This week's theme is resource allocation.
    • Key questions include:
    • How to share resources fairly?
    • How to prevent a network from being overloaded?
      • This leads to congestion control (discussed today).
    • How to ensure audio/video work well?
      • This relates to Quality of Service (QoS), discussed on Wednesday.

Key Topics to Cover

  1. What Congestion Is
  2. Basic Congestion Control Algorithms
    • Additive Increase, Multiplicative Decrease (AIMD).
  3. Calculating Resulting Speed
    • TCP Friendly formula.
  4. Refinements in Congestion Control
    • Slow start: Quickly ramp up connections.
    • Fast recovery: Maintain data flow after packet loss.
  5. Challenges in Long Fat Pipes
  6. Alternative Approaches
    • TCP Vegas: Detects congestion using delay.
    • TCP Cubic: Uses a third-power polynomial to manage the congestion window.

Understanding Congestion

  1. Congestion Definitions
    • Congestion occurs when too many packets are sent into the network at high rates, causing delays or packet loss.
  2. Congestion in a Single Queue
    • Example scenarios:
      • Regular packet arrivals lead to predictable delays.
      • Random packet arrivals cause varying delays.
  3. Finite Queue Implications
    • If the queue is full, packets are dropped, necessitating retransmission and extra work from end hosts and routers.
  4. Multiple Flows and Links
    • Congestion can be worsened by multiple data flows competing for the same resources.
  5. Consequences of Congestion
    • Long queues lead to considerable delays.
    • Queue overflows cause packet loss.
    • Wasted bandwidth for packets that must be retransmitted.
    • Without feedback, congestion can lead to a