Nandit Soparkar

Trading computation for bandwidth: reducing communication in distributed control systems using state estimators

Describes a new framework for distributed control systems in which estimators are used at each node to estimate the values of the outputs at the other nodes. The estimated values are then used to compute the control algorithms at each node. When the estimated value deviates from the true value by more than a pre-specified tolerance, the actual value is broadcast to the rest of the system; all of the estimators are then updated to the current value. By using the estimated values instead of true value at every node, a significant saving in the required bandwidth is achieved, allowing large-scale distributed control systems to be implemented effectively. The stability, performance, and expected communication frequency of the reduced communication system are analyzed in detail. Simulation and experimental results validating the effectiveness and communication savings of the framework are also presented.

Frequent Itemset Counting Across Multiple Tables

Available technology for mining data usually applies to centrally stored data (i.e., homogeneous, and in one single repository and schema). The few extensions to mining algorithms for decentralized data have largely been for load balancing. In this paper, we examine mining decentralized data for the task of finding frequent itemsets. In contrast to current techniques where data is first joined to form a single table, we exploit the inter-table foreign key relationships to obtain decentralized algorithms that execute concurrently on the separate tables, and thereafter, merge the results. In particular, for typical warehouse schema designs, our approach adapts standard algorithms, and works efficiently. We provide analyses and empirical validation for important cases to exhibit how our approach performs well. In doing so, we also compare two of our approaches in merging results from individual tables, and thereby, we exhibit certain memory vs I/O trade-offs that are inherent in merging of decentralized partial results.

A design methodology for distributed control systems to optimize performance in the presence of time delays

When a control system is implemented in a distributed fashion, with multiple processors communicating over a network, both the communication delays associated with the network and the computation delays associated with the processing time can degrade the systems performance. In this case, the performance of the system may depend not only on the performance of the individual components but also on their interaction and cooperation. The approach taken in this paper assumes that the control has been designed without taking into account the network architecture. A theoretical framework is presented which allows the effect of time delays on the mechanical performance of the system to be precisely modelled, and these models are used to determine the optimal network architecture for the given control system. A design example of a two-axis contouring system is presented.

Failure-resilient transaction management in multidatabase

Failure recovery in a multidatabase environment is addressed. It is shown that local autonomy considerations force the designer of a multidatabase system to trade off certain desirable properties to achieve reliability for transaction management. Representative techniques in the research literature are contrasted and compared. The authors approach to the problem is described.<<ETX>>

Data-valued partitioning and virtual messages (extended abstract)

Network Partition failures in traditional Distributed Databases cause severe problems for transaction processing. The only way to overcome the problems of “blocking” behavior for transaction processing in the event of such failures is, effectively, to execute them at single sites. A new approach to data representation and distribution is proposed and it is shown to be suitable for failure-prone environments. We propose techniques for transaction processing, concurrency control and recovery for the new representation. Several properties that arise as a result of these methods, such as non-blocking behavior, independent recovery and high availability, suggest that the techniques could be profitably implemented in a distributed environment.

The HANDS project: Studying and refining the automated collection of a cross-setting clinical data set

The consistent availability of a core set of clinical nursing data is essential to promote quality patient care. Although important work to improve terminology and enhance comparability of data is underway, the efforts do not address the immediate need for useful nursing data sets and valid methods of collection at the point of data entry. The Hands-on Automated Nursing Data System (HANDS) project is dedicated to refining a feasible methodology for gathering, storing, and retrieving a standardized nursing data set. To date the project team has developed and tested a prototype research tool that is automated and contains the structured terminologies (North American Nursing Diagnosis Association, Nursing Outcomes Classification, and Nursing Interventions Classification) to represent nursing diagnoses, outcomes, and interventions, respectively. The Phase I project development activities are reported in this article, along with Phase II and III plans for testing and refining the methodology under actual clinical conditions. Results and lessons learned during Phase I are reported.

Decentralized, modular real-time control for machining applications

Several different architectures for a control system for a reconfigurable machining system are considered, in terms of the inherent delays and their effects on the system performance. It is shown how time delays associated with distributed architectures impact the performance of the control algorithms, and how different types of communication protocols could be implemented to meet the required deadlines. A two-axis contouring system is considered in some detail, and the effects of various delays on the contour error are determined by simulation. These analytic and simulation results can be used to specify the maximum allowable communication delays in the system. Control requirements for a manufacturing control system can thus be mapped to temporal constraints on the data managed for the environment.

Systems for KDD: From concepts to practice

Abstract The considerable interest in knowledge discovery in databases (KDD) has led to several techniques and tools for the automated extraction of useful information from large data repositories. In order to use these developments in practical settings, there is need to consider the computing systems that would support the complete KDD process. In this regard, we identify and discuss important computing systems issues, and we compare and contrast some available research and commercial efforts. We suggest several enhancements to the underlying database systems that may significantly benefit the KDD process. We indicate why it is important that tools to handle different aspects of the KDD effort need to be integrated. Finally, we briefly describe our experience in implementing a prototype KDD system for a large corporate environment.

Probabilistic Data Consistency for Wide-Area Applications

Wide-area applications, such as electronic commerce, groupware, directory service etc., rely on distributed data management services which often employ data replication to meet availability and response time constraints imposed by the target domains. Since many of these applications can tolerate some staleness in the data they access, the constraints may be better met by exploiting relaxed data consistency among copies in the system. However, in the current Internet architecture, it is nearly impossible to provide deterministic guarantees for the bounds on data inconsistency among copies. We present an approach for providing probabilistic data access guarantees while maintaining availability and timing predictability. Our model makes few assumptions regarding the underlying infrastructure. In fact, a novel feature of our approach is that it is independent of the underlying replication scheme. We focus on modeling the update patterns of the servers and modifying the data access mechanisms. Our model guarantees, with a certain probability P , that if a client requests the value of an object, the value returned by the system, if any, will be temporally consistent with the newest copy of the same object in the system. That is, from the perspective of a client, the system provides a probabilistic data consistency guarantee.

Characterizing multicast orderings using concurrency control theory

Coordinating distributed executions is achieved by two widely used approaches: process groups and transactions. Typically, the two represent a trade-off in terms of the degrees of consistency and performance. By applying transaction concurrency control techniques to characterize and design process group multicast orderings, we aim to provide aspects of both ends of the trade-off. In particular, we propose a framework in which each message multicast is regarded as a transaction. Appropriate message ordering protocols are devised and shown to be correct using a variant of concurrency control theory. Also, we are able do incorporate certain aspects of application semantics for which existing process group approaches are inadequate. Finally, our framework provides a means to characterize the performance of orderings to allow a comparison of different ordering protocols.