Client-server and peer to peer

Overview

Recall questions on client-server and peer to peer networks, how each works, when each is the right choice, why peer to peer downloads speed up with more users, and evaluating a move from peer to peer to client-server.

What you need to know

Answer each one before you reveal it.

Questions from this subtopic

Each one with the model answer underneath.

1. 1

   Describe how a client-server network works.

   Model answer

   Machines take one of two distinct roles. Servers are dedicated, usually powerful machines that hold shared resources - files, databases, web pages, print queues, user accounts - and wait for requests. Clients are the users' machines; they initiate every exchange by sending a request to a server, and the server processes it and returns a response. Clients do not communicate with one another directly for these services, and the server does not initiate contact. Administration is centralised: accounts, permissions, security policy, software updates and backups are managed on the servers.

2. 2

   Describe how a peer to peer network works.

   Model answer

   There is no dedicated server and no fixed hierarchy. Every machine is a peer of equal status and acts as both client and server: it requests resources from other peers and simultaneously makes its own resources - files, storage, printers, processing power, upload bandwidth - available to them. Peers connect to one another directly, and each machine's own user controls what it shares and who may access it, so administration is distributed rather than central.

3. 3

   Compare client-server and peer to peer networks.

   Model answer

   Administration: client-server is managed centrally from the servers, so policy, accounts and updates are applied in one place; peer to peer must be configured machine by machine. Security: client-server can enforce consistent authentication and permissions centrally, whereas in peer to peer security is only as good as the least careful user's settings. Backup: client-server data sits on servers and one backup regime covers everyone; in peer to peer data is scattered across machines that may be switched off. Cost: client-server needs dedicated server hardware, server software and specialist administration; peer to peer needs none of these and uses machines that already exist. Reliability: a client-server network has a single point of failure - if the server goes down nobody can work; a peer to peer network keeps going when individual peers leave, though a resource is unavailable whenever the machine holding it is off. Performance and scalability: in client-server, every extra client adds load to the same server, so performance falls as the network grows unless the server is upgraded; in peer to peer, each extra peer brings its own capacity, so the system can grow without a central upgrade, but performance for any one transfer depends on the peers involved.

4. 4

   When is client-server the right choice?

   Model answer

   When central control matters: an organisation that must guarantee who can see which data, apply a single security policy, keep an audit trail, or comply with data protection obligations. When data must be reliably backed up and consistent, so that everyone works on one authoritative copy - a shared database, a school's records. When the service must be available on demand rather than depending on which users happen to be online. When resources are too large or too specialised to sit on user machines. And whenever the network is big enough that configuring every machine individually would be unmanageable.

5. 5

   When is peer to peer the right choice?

   Model answer

   For a very small network - a handful of machines in a home or a tiny office - where a dedicated server and an administrator cannot be justified for the sake of sharing a printer and a few files. For distributing large files to very many recipients, where letting each downloader also upload to others spreads the load instead of concentrating it on one server. For ad hoc, temporary sharing between people who happen to be present, with no infrastructure to set up. And where resilience against the loss of any single machine matters more than central control, since there is no central component to fail or to be attacked.

6. 6

   Why does distributing a large file peer to peer get faster as more people download it, when a client-server download gets slower?

   Model answer

   With a single server, every downloader draws on the same finite upload bandwidth and the same machine's processing capacity, so the available capacity is divided among them: twice as many downloaders means roughly half the speed each. In peer to peer distribution, each downloader receives pieces of the file and immediately makes the pieces it already has available to others, so every new participant adds upload capacity as well as demand. The total capacity therefore grows with the number of participants rather than staying fixed, and a peer can pull different pieces from several sources at once. The trade-offs are that the downloader's own upload bandwidth is consumed, that speed depends on how many peers are online and how good their connections are, and that there is no single authority guaranteeing that what is shared is genuine.

Other ways to work this

The same content, four other ways.

What is inside Levo

Six ways to work the same material.