Dec 1998 NI222: Networks and the Internet (Question 5)

(a)	Briefly explain how, when sending out an IP packet, the destination Ethernet address is determined.	[6]
	ARP (Address Resolution Protocol) is used to translate IP address to Ethernet addresses (1 mark). The translation is done only for outgoing IP packets, when the IP header and the Ethernet header are created (1 mark). The translation is performed with a look-up table called the ARP table ( 1 mark). This is stored in memory and relates the IP address with the ethernet address (1 mark). If the relevant IP address is not found in the ARP table, then a request is broadcast on the local network (1 mark), and the host with the requested IP address (or the router forwarding packets towards that IP address) will respond with its ethernet address (1 mark).
(b)	A modern approach to LAN construction uses a LAN hub. Explain what this means, and how this topology has elements of a star form and a bus form.	[3]
	Each device on the network is directly linked to a central hub (1 mark). Within this hub there is a high-speed bus on which a contention protocol operate (1 mark), and this is used to facilitate communication between the devices. It is a star in that each node is connected to a central node (the hub), and a bus in that the actual communication is via a bus at the central node (1 mark).
(c)	What is the major modern competitor to high-speed ethernets ? How do the capacities compare ? Describe one problem with this alternative, and one benefit.	[5]
	The alternative is Asynchronous Transfer Mode (ATM) (1 mark). Current ATM networks have speeds in excess of 500 Mbps (1 mark), which compare favourably with all but the gigabit ethernet. ATM is unlike IP, and so it is hard to run TCP-style connections over it efficiently (1 mark). ATM is well-optimised to carry audio or video transmission, because it is less vulnerable to network congestion (1 mark). Allow credit for appropriate alternative answers.
(d)	Is it true that the shortest path between any two nodes must also be the most economical path ? Explain your answer.	[6]
	No (1 mark). Shortest path algorithms are conceptually simple and widely used (1 mark). The advantage is the packets move along the most efficient path (1 mark). The main disadvantages are that they are not robust to the failure of links and changes to the cost of links (1 mark)—if links break, packets get stuck until shortest path is re-established (1 mark). They are also single path algorithms, hence heavy traffic can cause congestion (1 mark). Allow credit for appropriate alternative answers.