In process computation metrics provides a framework to compare and evaluate interconnection networks. About PRAM interconnection model there have following metrics that makes a difference among various PRAM interconnection networks;
Also Check: PRAM Interconnection Network model
Node Degree or Processor:
The number of links or communicating channels incident on node/processor (i.e., either processor node or memory module) is called the node degree or the number of neighbor processor. It reflects the number of I/O ports required per node. In case of using switches, it is also called the degree of the switch.
Network Degree:
The degree of the network is the maximum degree of any node/processor of the network.
Diameter:
The diameter D is the shortest path between any two nodes and the path length is measured in terms of links traversed.
Network Diameter:
The diameter of the network is the maximum distance between any two processors of the network. Where the distance between any two processors can be defined in the standard graph theoretic way, i.e., the distance from Pi to Pk is the minimum d for which there exists a sequence Pi, = Pi0, Pi1,. . . Pid = Pk, where P, is directly connected to (or neighbor of) Pi,j+1, 0 ≤ j < d. From the communication point of view, diameter should be kept small.
Note: Any degree k network has diameter at least logk(p -1). A network is of bounded-degree if its degree is bounded. Hence any bounded-degree network with p processors must have diameter Ω(log p).
Bisection width:
The smallest number of links cut from the network to disconnect the network into two equal halves known as the bisection of the width. It is also named as the channel bisection. Moreover, the number of bits that an be communicated simultaneously over a link connecting two nodes is called channel width. Channel width is equal to the number of physical wire in each communication link. The peak rate at which a single wire can deliver bits is called the channel rate. The peak rate at which data can be communicated between the ends of a communicating link is called channel bandwidth. It is the product of channel rate and channel network is referred as the bisection bandwidth of a network. It is the product of the bisection width and the channel bandwidth.
Network Connectivity:
Network connectivity measures its ability to continue operation despite disabled components. It is defined as the minimum number of nodes or links that must fail to partition the network into two or more disjoint networks.
Number of edges per node:
If each processor or nodes in the interconnection network have constant numbers links that are independent of the network size will be best for network, because the processor organization scales more easily to systems with large number of nodes.
Data routing function:
A data-routing network is used for data exchange among processing nodes or memory modules. This routing network can be static, such as hypercube routing networks. In case of multicomputer, the data routing is achieved through message passing. The versatility of a routing network will reduce the time for data exchange and thus significantly improve the system performance. Common data routing functions includes shifting, rotation, permutation, broadcast, multicast, shuffle exchange etc.
Cost:
The number of linking wires or communication links required by a network used to evaluate the cost of a network.
Interconnection Network Performance:
The performance of an interconnection network is affected by the following factors:
- Functionality: This refers to how the network support data routing, interrupt, handling, synchronization, request/message combining and coherence.
- Network Latency: This refers to the worst-case time delay for a unit message to be transferred through the network.
- Bandwidth: This refers to the maximum data transfer rate in terms of Millions Bits Per Second (MBPS) transmitted through the network.
- Hardware complexity: This refers to implementation costs such as for wires, switches, connectors, arbitration and interface logic.
- Scalability: This refers to the ability of a network to b modularly expandable with a scalable performance with increasing machine resources.
- Network Throughout: It is defined as the total number of message the network can handle per unit time. Typically, the maximum throughput of a network is some fraction of its capacity.
