IB Computer Science Topic A2.2.3: Compare and Contrast Networking Models Notes

Client-Server and Peer-to-Peer: Where Control and Resources Sit

1. A networking model describes how the machines on a network divide up the work of supplying and requesting resources.
2. In the client-server model that division is fixed: some machines only supply and the rest only request.
3. In the peer-to-peer model every machine does both jobs, so there is no fixed supplier.

What the Two Models Have in Common

1. Both run over the same physical network and the same TCP/IP protocols, so the model is a software arrangement rather than a different kind of cable.
2. Both still involve one machine requesting and another supplying, and the difference is only whether that role is permanent.
3. Both have to solve the same problems: finding the other machine's address, checking who is allowed to connect, and transferring the data reliably.
4. Both can serve very large numbers of users, and neither is secure by default.

Client-Server Model: Central Control, Central Risk

1. How It Works
   1. A dedicated server holds the resource, and clients send requests to it and wait for a reply.
   2. Clients never talk to each other for that resource, so the server sees and controls every transaction.
2. Benefits
   1. Centralized control means one place to set permissions, apply updates and take backups.
   2. Security is easier to enforce, because there is one gate to defend rather than one per machine.
   3. Data consistency follows from having one authoritative copy, which is why banking cannot use a model where every device holds its own version of your balance.
   4. Efficiency comes from hardware built for the one job, such as a database server with fast storage and a lot of memory.
   5. Reliability is high in practice, because a small number of machines can be given redundant power, disks and network links.
3. Drawbacks
   1. The server is a single point of failure, so when it stops, every client loses the service at once.
   2. Cost is high, because server hardware, software licences and staff to run them all have to be paid for before the first client connects.
   3. Every extra client adds load without adding capacity, so growth is paid for by upgrading the server or buying more of them.

Peer-to-Peer Model: Every Node Is Both Client and Server

1. How It Works
   1. Each peer supplies resources to others at the same time as it requests resources for itself.
   2. There is no central authority, so peers have to discover each other and agree on what is true between themselves.
2. Benefits
   1. There is no single point of failure, so losing individual peers degrades the network instead of stopping it.
   2. Scalability works the opposite way round: each new peer brings its own bandwidth and storage, so a popular file downloads faster the more people want it.
   3. Cost is low, since there is no server infrastructure to buy or house.
   4. Setup is simple on a small network, which is why file and printer sharing between a few home machines is peer-to-peer.
3. Drawbacks
   1. Without central control there is nowhere to enforce a password policy, apply a patch or take one reliable backup.
   2. Performance varies with which peers happen to be online and how much bandwidth they are willing to give.
   3. Anyone can join and offer content, so peers can distribute malware or infringing material with no gatekeeper to stop them.
   4. Data consistency is hard, because two peers can hold different versions of the same file and nothing decides which is correct.

Where Each Model Is Used in the Real World

1. Web Browsing (client-server)
   1. The browser is the client and the web server holds the pages, so the site owner controls exactly what every visitor receives.
2. Email Services (client-server)
   1. A mail client connects to a mail server so that the same mailbox appears on a phone and a laptop, which needs one stored copy that both devices read.
3. Online Banking (client-server)
   1. Every transaction is processed on the bank's servers, because a balance has to have exactly one correct value and an auditable record of who changed it.
4. File Sharing (peer-to-peer)
   1. BitTorrent splits a file into pieces and each peer uploads the pieces it already holds while downloading the ones it does not.
   2. This is why a heavily shared file arrives faster than a rarely shared one, which reverses the usual relationship between demand and speed.
5. VoIP Services (peer-to-peer)
   1. Once a server has introduced two callers, the audio can flow directly between them, which cuts both the delay and the operator's bandwidth bill.
6. Blockchain (peer-to-peer)
   1. Every node keeps its own copy of the ledger and the nodes agree on new entries between themselves, so there is no company whose server could be seized or shut down.

Comparing the Two Models Side by Side