Chapter 6 Congestion Control and Resource Allocation

Transcription

1 Chapter 6 Congestion Control and Resource Allocation Overview of Congestion Control and Resource Allocation Problem: How to effectively and fairly allocate resources among a collection of competing users? Resources being shared include o Bandwidth of the links o Buffers on the routers and switches where packets are queued awaiting transmission Resource allocation is the process by which network nodes try to meet the competing demands that applications have for network resources o Part of the problem is deciding when to say no and to whom What is congestion? o Packets contend at a router for the use of a link, with each contending packet placed in a queue waiting for its turn to be transmitted over the link o When too many packets are contending for the same link, the queue overflows and packets get dropped o When drops become common events, the network is said to be congested Congestion control refers to the efforts made by network nodes to prevent or respond to overload conditions o Some sort of resource allocation is built into the congestion control mechanisms

2 Congestion control and resource allocation are closely related o If the network takes an active role in allocating resources, then congestion may be avoided and there is no need for congestion control But allocating resources with any precision is difficult because resources are distributed throughout the network o On the other hand, we can let the sources send as much data as they want and then recover from congestion when it occurs Easier approach but many packets may be dropped by the network before congestion can be controlled Congestion control and resource allocation involve both hosts and routers o In routers various queuing disciplines can be used to control the order in which packets get transmitted and which packets get dropped o At end hosts the congestion control mechanism paces how fast sources are allowed to send packets Network Model o We consider resource allocation in a packet-switched network (or internet) consisting of multiple links and switches (or routers) A source may have more than enough capacity on the immediate outgoing link to send a packet, but somewhere in the middle of a network, its packets encounter a congested router o Connectionless flows We assume that the network is connectionless, with any connection-oriented service implemented in the transport protocol The datagrams are switched independently, but it is usually the case that a stream of datagrams between a

3 particular pair of hosts flows through a particular set of routers A sequence of packets sent between a source/destination pair and following the same route through the network is called a flow Flows can be defined at different granularities, e.g., host-to-host or process-to-process Routers can allocate resources to flows o Service model We focus on the best-effort service model in which all packets are given equal treatment A service model that supports some kind of preferred service or guarantee is said to provide multiple qualities of service (QoS) Taxonomy of resource allocation mechanisms o Router-Centric versus Host-Centric In a router-centric design, each router decides when packets are forwarded and which packets are to be dropped, as well as informs the source hosts how many packets they are allowed to send In a host-centric design, the end hosts observe the network conditions (e.g., how many packets they are successfully getting through the network) and adjust their behavior accordingly Both the routers and the hosts participate in resource allocation, the real issue is where the majority of the burden falls o Reservation-Based versus Feedback-Based In a reservation-based system, some entity (e.g., the end host) asks the network for a certain amount of capacity to be allocated for a flow

4 Each router allocates enough resources (buffers and/or percentage of the link s bandwidth) to satisfy the request If the request cannot be satisfied at some router, then the router rejects the reservation In a feedback-based approach, the end hosts begin sending data without first reserving any capacity and then adjust their sending rate according to the feedback they receive Explicit feedback: a congested router sends a please slow down message to the host Implicit feedback: the end host adjusts its sending rate according to the externally observable behavior of the network, such as packet losses A reservation-based system always implies a routercentric mechanism A feedback-based system can imply either a routercentric or host-centric mechanism If the feedback is explicit, then the router is involved in the resource allocation scheme If the feedback is implicit, then almost all burden falls to the end host o Window-Based versus Rate-Based Resource allocation mechanisms need a way to express to the sender how much data it is allowed to transmit Window-based mechanism uses a window to support resource allocation within the network Rate-based mechanism controls sender s behavior using a rate, that is, how many bits per second the network is able to absorb

5 o Best-effort service model usually implies feedback-based, hostcentric, window-based resource allocation mechanism o QoS-based service model usually implies reservation-based, router-centric, rate-based resource allocation mechanism Evaluation Criteria: how to determine if a resource allocation mechanism is good or bad? o Effective resource allocation We want as much throughput and as little delay as possible. Unfortunately, these two goals are often somewhat at odds with each other The power of the network is a metric for evaluating the effectiveness of a resource allocation scheme Power = Throughput/Delay Power is a function of the network load. The load, in turn, is set by the resource allocation mechanism o Fair resource allocation Assuming that fair implies equal share of bandwidth and that all paths are of equal length, Raj Jain proposed a metric that can be used to quantify the fairness of a resource allocation mechanism. Given a set of flow throughputs (x 1, x 2,..., x n ), Jain s fairness index is defined as The fairness index always results in a number between 0 and 1, with 1 representing greatest fairness Example 1: if all n flows receive the same throughput, then f = 1 Example 2: if k of the n flows receive equal throughput and the remaining flows receive zero throughput, then f = k/n