Neven Tomov

Local search and the number of solutions

There has been considerable research interest into the solubility phase transition, and its effect on search cost for backtracking algorithms. In this paper we show that a similar easy-hard-easy pattern occurs for local search, with search cost peaking at the phase transition. This is despite problems beyond the phase transition having fewer solutions, which intuitively should make the problems harder to solve. We examine the relationship between search cost and number of solutions at different points across the phase transition, for three different local search procedures, across two problem classes (CSP and SAT). Our findings show that there is a significant correlation, which changes as we move through the phase transition.

Approximate Estimation of Transaction Response Time

This paper describes a study of different approximation techniques used to predict the response times of database transactions represented as patterns of resource consumption and modelled with non-product-form queueing networks. The techniques are applied to a range of examples. The experiments show that none of the approximation techniques has a consistent advantage over the others for all cases considered. On the other hand, a simple heuristic rule is formulated which provides an acceptable approximation to the average transaction response time for the entire range of examples. The rule specifies a procedure for labelling each queue in a queueing network as either an M/M/1 or an M/G/1 resource. The resulting network can then be solved to obtain the mean response time of individual transactions.

An analytical tool for predicting the performance of parallel relational databases

The uptake of parallel DBMSs is being hampered by uncertainty about the impact on performance of porting database applications from sequential to parallel systems. The development of tools which aid the system manager or machine vendor could help to reduce this problem. This paper describes an analytical tool which determines the performance characteristics (in terms of throughput, resource utilisation and response time) of relational database transactions executing on particular machine configurations and provides simple graphical visualisations of these to enable users to obtain rapid insight into particular scenarios. The problems of handling different parallel DBMSs are illustrated with reference to three systems – Ingres, Informix and Oracle. A brief description is also given of two different approaches used to confirm the validity of the analytical approach on which the tool is based. Copyright

Cache modelling in a performance evaluator for parallel database systems

Cache modelling is an important issue in developing an analytical performance evaluator to estimate the performance of applications running on parallel DBMSs. This paper describes a cache model developed for parallel cache management in the Oracle7 parallel server. Some preliminary results have also been obtained by using the cache model to predict the cache hit ratio for varying database sizes and varying numbers of participating processing elements.

Verifying a Performance Estimator for Parallel DBMSs

Although database systems are a natural application for parallel machines, their uptake has been slower than anticipated. This problem can be alleviated to some extent by the development of tools to predict the performance of parallel database systems and provide the user with simple graphic visualisations of particular scenarios. However, in view of the complexities of these systems, verification of such tools can be very difficult. This paper describes how both process algebra and simulation are being used to verify the STEADY parallel DBMS performance estimator.

Modelling Parallel Oracle for Performance Prediction

The problem of predicting the performance of a parallel relational DBMS for a set of queries applied to a particular data set on a shared nothing parallel architecture without transferring the application to a parallel system is a challenging one. An analytical approach has been developed to assist with this task and has been applied to the ICL GoldRush machine, a parallel machine with a shared-nothing architecture. This paper describes how the Oracle Parallel Server and the Parallel Query Option are modelled by the method and compares the predictions of the model against actual measurements obtained.

Some Results from a New Technique for Response Time Estimation in Parallel DBMS

The need for tools for performance prediction of parallel database systems is generally recognised. One such tool which has been developed (Steady) is based on analytical techniques to obtain a rapid estimate of performance. The approach to predicting response time involves a heuristic approximation coupled with standard queueing solutions. This paper reports on preliminary results for both maximum transaction throughput and response time obtained in comparing this approach against actual measurements.

Decision Supporting for Management of Parallel Database Systems

Parallel database systems are generally recognised as one of the most important application areas for commercial parallel systems. However, the task of managing the performance of a parallel database system is exceedingly complex. The initial choice of hardware configuration to support a particular DBMS application and the subsequent task of tuning the DBMS to improve performance rely not only on the way in which the data is structured, but also on how it is fragmented, replicated and distributed across the processing elements of the system. To understand the behaviour of a particular application requires the study of large volumes of performance data. To simplify this process it is essential to provide some means of presenting performance data in a comprehensible form which will aid visualisation. This paper explores some of the issues relating to decision support for the performance management of parallel database systems and describes an analytical capacity planning tool to assist users in this task.

STEADY - A Tool for Predicting Performance of Parallel DBMSs

Predicting the performance of a parallel relational DBMS executing an arbitrary set of transactions on particular data sets for different architectural configurations with different data placement strategies is a non-trivial task. An analytical tool has been developed to assist with this task and can be used for application sizing, capacity planning and performance tuning.