Skip to main content

9. Physical and Architectural Models

07/03/23

MoodlePDF

  • Physical models - Hardware components and networks
  • Architectural models - Computation and communication tasks and their relationships
  • Fundamental models - Abstract, specialised models
    • Interaction models
    • Failure models
    • Security models

Physical Models​

Generations of distributed systems​

  • Early - Late 1970s/early 1980s, 10-100 nodes on a local area network
  • Internet Scale - 1990s/early 2000s, many nodes but mainly desktop/server computers
  • Contemporary - Late 2000s+, plus mobile devices, ubiquitous computing/IoT (embedded devices) and cloud computing

Architectural Models​

1. Communicating entities​

From a system perspective

  • Communicating entities are usually processess managed by operating system
  • Simple computing nodes may have no operating system

From a programming perspective

  • Objects: as in Object-Oriented programming, present well-defined interfaces made up of methods
  • Components: Have interfaces similar to objects, also describe their dependencies
  • Web Services: With interfaces defined in terms of WWW technologies and standards

2. Communication Paradigms​

From system perspective

  • Message-passing primitives
  • Sockets interface for messages and streams (UDP, TCPIP)
  • Multicast (1-M) physical communication
  • Direct Communication = Message exchange between specific identified (and addressed) entities
    • Remote invocation - direct imperative-style message exchange
  • Request-reply protocols - Client server stype (HTTP)
  • Remote Procedure Call (RPC) and Remote Method Invocation (RMI)
    • make remote communication like local code execution
  • Internet Communication - where a message is not addressed to one specific entity

3. Roles and Responsibilities​

Client Server

  • Cleint initaites requests
  • Server handles requests against the resources that it manages and responds with a result
  • A server can be a client for other servers

Peer-to-peer

  • No distinction between entities in the network - every node or peer is essentially the same
  • Each node typically runs the same program, and manages some local resources

4. Placement​

Can affect: performance, reliability, security, scalability

Common strategies​

  • Mapping of a service of multiple servers - several processes on different machines can work together to provide a single logical service (WWW, DNS)
  • Caching - Temporarily storing recently used data closer to where it was
  • Mobile Code - Moving code to another process to run rather than moving data
  • Mobile agents - Moving complete running processes to another node

Architectural Patterns​

Ways of organising distributed systems that are found to be useful in many different systems

Layering​

  • Partitioned into a number of layers, with a given layer making use of the services offered by the layer below
  • E.g. network protocol stacks

Tiered Architectures​

Functions of the system are divided into distinct layers or tiers

  • Each of which is comprised of specific processes e.g.
  • Two-tier client server system, distributed printing, thin-client/remote desktop
  • Three-tier web system, with separate (dedicated) database server, web (application) servers and clients (browsers)

Some more patterns​

  • Caching
  • Proxy - An object (or service) presents the same interface as a remote service but more locally, and passes on requests (perhaps with some added functionality)
  • Broker - A common point where service providers can register their existence so that clients can find them

Middleware​

A common software layer, above the operating system but common to many applications, that supports particular communication paradigm(s) Tailored to mix and match