There are various forms of interconnection required in a parallel computation. The PRAM interconnection network model can play an important role as a theoretical yardstick for measuring the limits of parallel computation technique. This type of interconnection network models used to affects the way in which efficient algorithms are developed. In parallel computation the communication between its processors can be done trivially through its shared memory; these models can be also used to detect the intrinsic parallelism of a given problem. Moreover, the techniques and the paradigms provided by the PRAM algorithm can be utilized to design algorithms on more realistic models. However, the PRAM model is not easy to realize physically because of physical fan-in limitations. In a physically realizable assemblage, we can only expect any computing element to have a small number of external connections. We must therefore consider parallel assemblages in which a large number of communicating processors, each with its own memory, are connected together, but where each processor communicates with a small number of other Processors.
Figure: The Network Model
The mechanism for data transfer between processing nodes or between processors and memory modules
An interconnection networks provide the mechanisms for data transformation in between processing nodes and/or processors and/or memory modules in parallel computation over communication. Typical interconnection networks are built by using links and switches. A link corresponds to physical media such as a set of wires or fibers. These links are capable to operate processors synchronously and communicate them with one another by sending and receiving data packets over these communication links provided by the network. Each processor can set up a single packet of bounded length in a unit of time. While switches are devices that unable multiple processors input their data to a specific output processor use. The communications of p-processor over PRAM models can be classified into two interconnection network models are as follows;
- Static PRAM interconnection networks models and
- Dynamic PRAM interconnection networks models
Figure: Classification of Interconnection network
Static Interconnection network:
Static interconnection network consist point-to-point communication links among processing nodes that will not change ones they are created and it is also referred to as direct networks. This interconnection network is used for fixed connection among subsystems of centralized system or multiple computing nodes of a distributed system. A p-processor static interconnection or fixed interconnection network may be viewed as an undirected graph, where vertices correspond to the processors and edges correspond to communication links among processors. These interconnection networks are more suitable for building computers where the communication patterns are predictable and implementable with static connections. Static interconnection network is further divided in to two categories, Regular (It is mostly used to interconnect loosely coupled processors.) and irregular (It is tightly coupled interconnection process).
Figure: Static Interconnection Network
Above figure illustrates a simple static network of four processing elements or nodes. Each processing node is connected via a network interface to two other nodes in a mesh configuration.
Examples of static interconnection network are Linear array, Ring, Mesh, Tree, Hypercube, Cube connected, Butterfly etc. Shared paths in static interconnection networks are formed by interconnecting processing elements with buses.
Dynamic Interconnection network:
Dynamic interconnection networks are built by using switches and the communication links among processors. The communication links are connected to one another dynamically by the switches to establish a desired path among processing nodes and memory banks. Dynamic networks are also referred to as indirect networks.
Figure: Dynamic Interconnection Network
Above figure illustrates a dynamic network of four nodes connected via a network of switches to other nodes. Dynamic connections can implement all communication patterns based on program demands for multipurpose or general purpose applications. Examples of dynamic interconnection network are Single state networks, Bus, Cross Bar Network, Multistage network, Omega network etc.
Re commanded Tutorial for You: