WEEK 4 CS2005 Distributed Systems Concepts

Session Objectives

• Understand the main principles of distributed systems

• Learn about communication in client-server systems

• Learn the concepts of socket programming

• Understand Remote Procedure Calls (RPCs)

• Explore the use of pipes for interprocess communication

Main Principles of Distributed Systems

• Definition: A distributed system consists of a collection of loosely coupled nodes interconnected by a communication network.

• Each node sees its own resources as local while other nodes' resources are seen as remote.

• Nodes: Vary in size and function; can include microprocessors, personal computers, and large computer systems.

• Terminology: Nodes can be referred to as processors, sites, machines, or hosts, with 'site' indicating location and 'node' indicating the specific system.

Client-Server Model

• Typically, one node (the server) has a resource that another node (the client) wishes to access.

Why Do We Need Distributed Systems?

• Resource Sharing: Users can access resources from different sites (e.g., printing documents located remotely).

• Computation Speedup: Distributing sub-computations allows for concurrent execution and load balancing across sites.

• Reliability: Failure of one site does not affect the operation of the rest, enhancing overall system reliability through redundancy.

• Communication: Enables information exchange between users and nodes, facilitating activities such as file transfers and remote procedure calls (RPCs).

Advantages and Disadvantages of Distributed Systems

Advantages

• Reliability: High fault tolerance

• Scalability: Can scale with increased resources

• Flexibility: Adaptable to various configurations and needs

• Speedup: Multiple processes can run simultaneously

• Openness: Can integrate heterogeneous systems

• High Performance: Improved speed and efficiency in processing

Disadvantages

• Difficult Troubleshooting: Harder to diagnose problems

• Less Software Support: Limited resources for distributed applications

• High Network Infrastructure Costs: Expensive to set up robust networks

• Security Issues: Increased vulnerabilities across the system

Processes Communicating

• A program during execution is termed a process.

• Inter-process Communication (IPC): Defined by the OS for processes running on the same host.

• For processes on different hosts, they communicate through message passing across a communication network.

• Roles: The initiating process is the client; the waiting process is the server.

Sockets

• Definition: Sockets allow client and server processes to communicate by reading from and writing to them, functioning similarly to doors.

• Communication Process:

  • Messages sent from a sending process through a socket are delivered to the receiving process's socket via the network.

  • The destination address includes both the IP address and port number for routing.

Addressing

• Destination Address: Consists of the destination host's IP and the port number of the receiving application.

• Source Address: Comprises the source host IP and source port number, typically managed by the OS.

Socket Programming

• Types:

  • UDP (User Datagram Protocol): Connectionless, unreliable, message-oriented.

  • TCP (Transmission Control Protocol): Connection-oriented, reliable, byte-stream oriented.

Socket Programming Examples

1. UDP Client-Server:

   • No prior connection establishment.

   • Each packet sent contains attached IP and port details.

   • Data can be lost or received out-of-order.

2. TCP Client-Server:

   • Client contacts the server which must be actively running.

   • TCP provides reliable communication, supporting multiple clients by creating new sockets for each connection.

Python Examples

• UDP Client/Server:

  • Client sends a message, server receives and modifies it, sending it back.

• TCP Client/Server:

  • Similar communication using the TCP protocol ensuring order and reliability.

Java Examples

• EchoServer and Single Client Knock-Knock Application demonstrate client-server interactions with TCP using echoing messages or jokes.

Remote Procedure Calls (RPCs)

• Definition: RPCs enable a client to invoke procedures on remote hosts as if they are local calls.

• Mechanism: Involves stubs on both client and server to manage communication details, including parameter marshaling and message transmission.

Pipes

• Purpose: Channels for interprocess communication.

Types

1. Ordinary Pipes: Unidirectional, requiring a parent-child relationship.

2. Named Pipes: Bidirectional, allowing multiple processes to communicate, even over networks.

Implementation

• Ordinary pipes in UNIX use pipe(int fd[]), where fd[0] is the read-end and fd[1] is the write-end.