Narrow Results

The Communication Machine brings to the multicomputer what vectorization brought to the uniprocessor. It provides the same tools to speed communication that have traditionally been used to speed computation; namely, the capability to program optimal communication algorithms on an architecture that can, to the extent possible, replicate their performance in terms of wall-clock time. In addition to the usual complement of logic and arithmetic units, each module contains a programmable communication unit that orchestrates traffic between the network and registers that communicate directly with comparable registers in neighboring modules. Communication tasks are performed out of these registers like computational tasks on a vector uniprocessor. The architecture is balanced in the sense that, on average, the speed of local and global memory is comparable. Theoretical performance is tabulated for both hypercube and mesh interconnection networks. The Communication Machine returns to the somewhat beleaguered, yet intuitive concept that the performance we ultimately seek must come from a truly massive number of processors.

The adaptive routing approach has been expected as a promising way to improve network performance by utilizing available network bandwidth. Previous adaptive routing strategies in wormhole-routed multicomputer networks restrict the routing of messages by the routing algorithm to prevent deadlock. This results in low degree of adaptivity and low utilization of physical or virtual channels. In this paper, we examine the possibility of performing restriction-free adaptive routing in wormhole-routed networks as an approach to further improving the performance of these networks. A new flow control policy, called message cutting-in, is proposed, and two adaptive routing strategies are presented. Freedom of communication deadlock is achieved by the proposed flow control policy. The proposed adaptive routing strategies do not restrict routing and maximally utilize the physical and virtual channels. Simulation results show that the restriction-free adaptive routing approach is promising from the fact that it has the lowest latency and highest throughput depending on the number of virtual channels per physical channel and patterns of message traffic.

In this paper, we study wormhole routed networks and envision their suitability for real-time traffic in a priority-driven paradigm. A traditional blocking flow control in wormhole routing may lead to a priority inversion in the sense that high priority packets are blocked by low priority packets for unlimited time. The priority inversion causes the frequent deadline missing even at a low network load. This paper therefore proposes two preemptive flow control policies where high priority packets can preempt network resources held by low priority packets. As a result, the proposed flow controls can resolve the priority inversion. Our simulations show that preemptive flow controls significantly reduce deadline miss ratios for various real-time traffic configurations.

For real-time communication, we must be able to guarantee timely delivery of messages. In a previous paper, Kim et al. presented a real-time communication method for networks which uses a deterministic wormhole routing algorithm. It would be more desirable to be able to use an adaptive wormhole routing algorithm. However, the use of an adaptive algorithm results in highly unpredictable communication delays because the path used by each message cannot be known in advance. Thus, an alternative is to use a flexible wormhole routing algorithm, in which one of a set of predefined paths is chosen (in advance) for each pair of communicating nodes. With flexible routing, real-time communication guarantees are again possible while making more effective use of the available network resources than deterministic routing. This paper examines the problem of selecting a set of paths to maximize the probability of meeting real-time communication guarantees for a set of communicating nodes. Since this problem is NP-hard, a heuristic solution is proposed and compared with previous path selection algorithms. Simulation results are used to show that the proposed path selection algorithm outperforms all previous algorithms.

Gossiping is proposed as the preferred communication primitive for replicating pivot data in dense matrix factorization on message passing multicomputer. Performance gains are demonstrated on a hypercube for LU factorization algorithms based on gossiping as opposed to broadcasting. This finding has consequences for the design of numerical software libraries.

We investigate the average-case scalability of parallel algorithms executing on multicomputer systems whose static networks are k-ary d-cubes. Our performance metrics are isoefficiency function and isospeed scalability. For the purpose of average-case performance analysis, we formally define the concepts of average-case isoefficiency function and average-case isospeed scalability. By modeling parallel algorithms on multicomputers using task interaction graphs, we are mainly interested in the effects of communication overhead and load imbalance on the performance of parallel computations. We focus on the topology of static networks whose limited connectivities are constraints to high performance. In our probabilistic model, task computation and communication times are treated as random variables, so that we can analyze the average-case performance of parallel computations. We derive the expected parallel execution time on symmetric static networks and apply the result to k-ary d-cubes. We characterize the maximum tolerable communication overhead such that constant average-case efficiency and average-case average-speed could be maintained and that the number of tasks has a growth rate Θ(P log P), where P is the number of processors. It is found that the scalability of a parallel computation is essentially determined by the topology of a static network, i.e., the architecture of a parallel computer system. We also argue that under our probabilistic model, the number of tasks should grow at least in the rate of Θ(P log P), so that constant average-case efficiency and average-speed can be maintained.