Trace-driven Simulation Models Research Articles

The process-flow model

Input/output subsystem performance is currently receiving considerable research attention. Significant effort has been focused on reducing average I/O response times and increasing throughput for a given workload. This work has resulted in tremendous advances in I/O subsystem performance. It is unclear, however, how these improvements will be reflected in overall system performance. The central problem lies in the fact that the current method of study tends to treat all I/O requests aa equally important. We introduce a three class taxonomy of I/O requests based on their effects on system performance. We denote the three classes time-critical, time-limited, and time-noncritical . A system-level, trace-driven simulation model has been developed for the purpose of studying disk scheduling algorithms. By incorporating knowledge of I/O classes, algorithms tuned for system performance rather than I/O subsystem performance may be developed. Traditional I/O subsystem simulators would rate such algorithms unfavorably because they produce suboptimal subsystem performance. By studying the I/O subsystem via global, system-level simulation, one can more easily identify changes that will improve overall system performance.

AN EVALUATION OF THE MEMORY REFERENCE BEHAVIOR OF ENGINEERING/SCIENTIFIC APPLICATIONS IN PARALLEL SYSTEMS

This paper presents the results of a study conducted to evaluate the inherent memory reference behavior of several engineering/scientific applications, executing on shared memory, MIN-based, parallel systems. In this study, system sizes of two to 64 processors were evaluated. A trace-driven simulation model was used to obtain dynamic reference characteristics of the code. Included in this code were explicit declarations of shared variables. Our results indicate that a significant amount of explicitly declared shared data is accessed as either readonly by several processors, or read-write by a single processor. Furthermore, lines containing synchronization variables tend to see small ownership times at a processor and are accessed by several processors in the system. We also note that, as expected, relatively more references are to data with smaller ownership times, as the number of processors increase. Finally, the application data set size can have an impact on ownership time, as the number of processors increase.

Evaluation of a multiple version scheme for concurrency control

This paper presents a thorough investigation of all relevant properties of a multiple version scheme for concurrency control. It offers conflict-free scheduling for reader transactions thereby generally reducing resource contention. By using a trace-driven simulation model we explored the effective parallelism achievable in multi-user database systems and the number of occurring deadlocks as a complementary measure. The workload was represented by six real-life object reference strings of different applications running on databases which vary in size between 60 MB and 2.9 GB. To compare and valuate the outcome of our experiments we used known results for RX-, RAC- and optimistic synchronization protocols derived under the same conditions. Furthermore, version-dependent criteria and their influence on general database performance were considered. The results obtained underline the value of the multiple version concept for concurrency control; many criteria vote for its use in future database systems with highly concurrent transactions.

Run-time characteristics of a simulation model

This paper compares the efficiency of trace-driven simulation models of computer systems with the efficiency of function driven models. In a trace-driven model (TDM), the next service request for a task is selected from a file of requests which was created earlier, often from data derived from a functioning system. This can be compared to a function-driven model (FDM), in which the next service request is selected using generation techniques based on random deviates. While both kinds of models can produce statistically accurate results, a TDM is thought to be inherently more accurate, but it has been suspected that a FDM is faster. The comparison of these two techniques is based on a simulation model of a computer system which has been specially constructed so as to operate either as a TDM or an FDM. The key feature of the study is that by using this model, we have isolated the demand-generation technique (either trace-driven or function-driven) as the single factor affecting performance of the model. A run-time program monitor and a software-monitor event probe are used to evaluate the resources used by various runs of the two kinds of simulation models. The results indicate that, contrary to earlier expectations, the performance of both techniques is comparable. We also demonstrate the effects of various block sizes on the performance of the TDM and the effects of the complexity of the request functions on the performance of the FDM.