Balancing consistency and lag in transaction-based computational steering

Abstract

Computational steering, the interactive adjustment of application parameters and allocation of resources, is a promising technique for higher-productivity simulation, finer-grained optimization of dynamically varying algorithms, and greater understanding of program behavior and the characteristics of data sets and solution spaces. Tools for computational steering must provide monitoring, visualization, and interaction facilities. In addition, these tools must address issues related to the consistency, latency, and scalability at each of these phases, and must consider the perturbation that results. In this paper we describe transaction-based components for a computational steering system and present an approach that guarantees consistent monitoring and displays,... Read complete abstract on page 2.Read complete abstract on page 2.Complete Abstract: Complete Abstract:Computational steering, the interactive adjustment of application parameters and allocation of resources, is a promising technique for higher-productivity simulation, finer-grained optimization of dynamically varying algorithms, and greater understanding of program behavior and the characteristics of data sets and solution spaces. Tools for computational steering must provide monitoring, visualization, and interaction facilities. In addition, these tools must address issues related to the consistency, latency, and scalability at each of these phases, and must consider the perturbation that results. In this paper we describe transaction-based components for a computational steering system and present an approach that guarantees consistent monitoring and displays, supports scalable monitoring, and provides the user with the ability to adjust the tradeoffs among lag, consistency and perturbation.

Creating an accurate portrayal of concurrent executions

“…The related sidebar explains our goals and achievements with this visualization environment and illustrates our practical support for reorderable, synchronous, and indepen-dent displays. The ideas and techniques developed through our work with Parade serve as the basis for continuing work with causality filters 5 and with visualization for the monitoring and control of "metacomputations," 6 parallel and distributed systems, 7,8 and network management. 9 Many debuggers, performance-evaluation systems, and program-understanding tools have incorporated visualization and animation into their presentations in order to easily convey complex information.…”

On-the-fly calculation and verification of consistent steering transactions

“…This is a nontrivial process in a distributed, asynchronous environment in which centralized control of the computation is required. This model can result in considerable perturbation of the computation and significant steering lag [8].…”