Dipak Ghosal

Approximate analysis of single and multiple ring networks

Asynchronous packet-switched interconnection networks with decentralized control are very appropriate for multiprocessing and data-flow architectures. The authors present performance models of single- and multiple-ring networks based on token-ring, slotted-ring, and register-insertion-ring protocols. The multiple ring networks have the advantage of being reliable, expandable, and cost effective. An approximate and uniform analysis, based on the gate M/G/1 queuing model, has been developed to evaluate the performance of both existing single-ring networks and the proposed multiple-ring networks. Approximations are good for low and medium load. The analyses are based on symmetric ring structure with nonexhaustive service policy and infinite queue length at each station. They essentially involve modeling of queues with single- and multiple-walking servers. The results obtained from the analytical models are compared to those obtained from simulation. >

Performance evaluation of a dataflow architecture

The formulation and validation of an analytical approach for the performance evaluation of the Manchester dataflow computer is discussed. The analytical approach is based on closed queuing network models. The average parallelism of the dataflow graph being executed on the dataflow architecture is shown to be related to the population of the closed network. The model of the dataflow computer is validated by comparing the analytical results to those obtained from the prototype Manchester dataflow computer and from simulation. The bottleneck centers in the prototype machine have been identified through the model, and various architectural modifications have been investigated from performance considerations. >

Modeling of hierarchical distributed systems with fault-tolerance

Since each of the levels in a hierarchical system could have various characteristics, different fault-tolerant schemes could be appropriate at different levels. A stochastic Petri net (SPN) is used to investigate various fault-tolerant schemes in this context. The basic SPN is augmented by parameterized subnet primitives to model the fault-tolerant schemes. Both centralized and distributed fault-tolerant schemes are considered. The two schemes are investigated by considering the individual levels in a hierarchical system independently. In the case of distributed fault tolerance, two different checkpointing strategies are considered. The first scheme is called the arbitrary checkpointing strategy. Each process in this scheme does its checkpointing independently; thus, the domino effect may occur. The second scheme is called the planned strategy. Here, process checkpointing is constrained to ensure no domino effect. The results show that, under certain conditions, an arbitrary checkpointing strategy can perform better than a planned strategy. The effect of integration on the fault-tolerant strategies of the various levels of a hierarchy are studied. >

Analytical modeling and architectural modifications of a dataflow computer

Dataflow computers are an alternative to the von Neumann architectures and are capable of exploiting large amount of parallelism inherent in many computer applications. This paper deals with the performance analysis of the Manchester dataflow computer based on queueing network models. The model of the dataflow computer has been validated by comparing the analytical results with those obtained from the prototype Manchester dataflow computer. The bottleneck centers in the prototype machine have been identified through the model and various architectural modifications have been investigated both from performance and reliability viewpoints.

Analysis Of Computation-Communication Issues In Dynamic Dataflow Architectures

This paper presents analytical results of computation-communication issues in dynamic dataflow architectures. The study is based on a generalized architecture which encompasses all the features of the proposed dynamic dataflow architectures. Based on the idea of characterizing dataflow graphs by their average parallelism, a queueing network model of the architecture is developed. Since the queueing network violates properties required for product from solution, a few approximations have been used. These approximations yield a multi-chain closed queueing network in which the population of each chain is related to the average parallelism of the dataflow graph executed in the architecture. Based on the model, we are able to study the effect on the performance of the system due to factors such as scalability, coarse grain vs. fine grain parallelism, degree of decentralized scheduling of dataflow instructions, and locality.

Trade-offs between response times and availability in a distributed database

Our goal is to study the conflicting interaction between two key performance measures in a replicated distributed database. The database is assumed to be prone to site and link failures, possibly leading to a network partition. The performance measures we consider are the availability of transactions, by which we mean the proportion of transactions ezpected to complete over a period of time [COK86], and the expected transaction response time degradation, by which we mean the average extra time needed to complete a transaction due to failures. The intuition behind this conflicting interaction is as follows. To increase availability in presence of failures, one would design a replica management protocol that expects the failed components to recover and participate in transaction executions. To design for such recovery inherently implies a degradation in response time characteristics, since we must wait for the failed components to recover. On the other hand, if response time is important (for example, there are hard real-time constraints on transactions), it is clear that fewer transactions would successfully complete; the rest prematurely aborted because of insufficient progress within the deadline. This degrades the availability of transactions. We believe that this is a very important conflict to analyze since it can be effectively used to