Hence we will focus on how switches can solve this

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: the switches to do the hard work of finding a usable set of paths. Hence, we will focus on how switches can solve this problem. Clearly, because the information required for solving the problem is spread across different switches, the solution involves the switches cooperating with each other. Such methods are examples of distributed computation. Our solution will be in three parts: first, we need a way to name the different nodes in the network. This task is called addressing. Second, given a packet with the name of a destination in its header we need a way for a switch to send the packet on the correct outgoing link. This task is called forwarding. Finally, we need a way by which the switches can determine how to send a packet to any destination, should one arrive. This task is done in the background, and continuously, building and updating the data structures required for forwarding to work properly. This background task, which will occupy most of our time, is called routing. ￿ 18.2 Addressing and Forwarding Clearly, to send packets to some end point, we need a way to uniquely identify the end point. Such identifiers are examples of names, a concept commonly used in computer systems: names provide a handle that can be used to refer to various objects. In our context, we want to name end points and switches. We will use the term address to refer to the name of a switch or an end point. For our purposes, the only requirement is that addresses refer to end points and switches uniquely. In large networks, we will want to constrain how addresses are assigned, and distinguish between the unique identifier of a node and its addresses. The distinction will allow us to use an address to refer to each distinct network link (aka “interface”) available on a node; because a node may have multiple links connected to it, the unique name for a node is distinct from the addresses of its interfaces (if you have a computer with multiple active network interfaces, say a wireless link and an Ethernet...
View Full Document

This document was uploaded on 02/26/2014 for the course CS 6.02 at MIT.

Ask a homework question - tutors are online