1 / 7

Queueing Problem

to next router. (drop-tail) queue. server. packet losses. Queueing Problem. The performance of network systems rely on different delays. Propagation/processing/transmission/queueing delays Which delay is affected most by network congestion? Queueing delay.

glora
Télécharger la présentation

Queueing Problem

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. to next router (drop-tail) queue server packet losses Queueing Problem • The performance of network systems rely on different delays. • Propagation/processing/transmission/queueing delays • Which delay is affected most by network congestion? • Queueing delay • How can we control this queue behavior? • How can we control the arrival rate and service rate? • Do we know anything about queueing behavior? • Can we analyze its behavior more rigorously using probability?

  2. packet arrivals time Preliminaries • How do we define the arrival process? • How do we define the service process? • Our assumptions • The arrival process follows a Poisson process. • The service process follows a Poisson process. • In other words, the interarrival times follow an exponential distribution. • Our ultimate interests • Average queue length • Average waiting time

  3. Our Goals • How do we compute the average queue length? • Suppose that the queue has 5 packets with 0.5 probability, 10 packets with 0.3 probability, and 20 packets with 0.2 probability. • What is the average queue length? • Can you generalize this process? • How do we compute the average wait time? • Assume the same probability distribution as the above. • What is the total wait time for a packet arriving with 4 previous packets in the queue? • Can you generalize this process?

  4. to next router (drop-tail) queue server packet losses Queue Mangement • Without proper queue management, we may have the following problems: • Packet drops • Global synchronization • Bias against bursty traffic • How do we manage queue behavior to mitigate these problems?

  5. Pdrop No dropping or marking Drop with P=1 1 Randomly drop some Drop all Pmax Mark with P Linearly increasing From 0 to Pmax Qavg 0 Thmin Thmax Drop Probability P Average Queue Length Random Early Detection (RED) • Try to detect incipient congestion early to reduce the number of packet drops. • Use randomness to mitigate global synchronization and bias against bursty traffic.

  6. Active Queue Management • Why RED good? • Simple • Not require source cooperation. • Not require per-connection state management. • Can identify connections using a large share of bandwidth (misbehaving flows). • Is RED perfect? Of course not. It raised many related questions • Do we count bytes or packets in RED? • RED with drop preference? • Any implementation issues in RED? • Marking or dropping, and TCP? • How about packet loss rate and link utilization? • Is RED effective for Web traffic?

  7. ECN marked Router Source Dest ACKs With ECN Explicit Congestion Notification (ECN) • Problems with non-ECN-compatible equipment: • 2,151 of 24,030 web servers were not accessible to ECN-capable • clients (tests in December 2000 using TBIT[2]).

More Related