{"title":"A performance debugger for eliminating excess synchronization in shared-memory parallel programs","authors":"R. Rajamony, A. Cox","doi":"10.1109/MASCOT.1996.501026","DOIUrl":"https://doi.org/10.1109/MASCOT.1996.501026","url":null,"abstract":"We describe a new approach to performance debugging that focuses on the automatic detection of unnecessary or excessive synchronization. We have implemented a prototype performance debugger that implements this approach, reporting the excess synchronization back to the user at the source level. We describe our performance debugger and report some results from applying it to a suite of programs. In particular, we report the type and amount of excess synchronization found in each program and the effect that eliminating this synchronization had on the program's performance. In one case, eliminating the excess synchronization identified by the performance debugger reduced the execution time by 41%.","PeriodicalId":256311,"journal":{"name":"Proceedings of MASCOTS '96 - 4th International Workshop on Modeling, Analysis and Simulation of Computer and Telecommunication Systems","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1996-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126931801","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Quantitative methods for design of benchmark suites","authors":"J. Dujmovic","doi":"10.1109/MASCOT.1996.501011","DOIUrl":"https://doi.org/10.1109/MASCOT.1996.501011","url":null,"abstract":"We present black-box models of workload difference and we then to develop quantitative methods for the design of benchmark suites. The proposed design methods (the method of optimum subsets, and the method of optimum weights) combine a new concept of program space coverage functions, and the criteria of completeness, redundancy, density, and granularity of benchmark suites.","PeriodicalId":256311,"journal":{"name":"Proceedings of MASCOTS '96 - 4th International Workshop on Modeling, Analysis and Simulation of Computer and Telecommunication Systems","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1996-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128767999","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Using remote memory to avoid disk thrashing: a simulation study","authors":"E. Markatos","doi":"10.1109/MASCOT.1996.500996","DOIUrl":"https://doi.org/10.1109/MASCOT.1996.500996","url":null,"abstract":"The increasing use of high-bandwidth and low-latency networks make possible the use of remote (network) memory as an alternative to disk means of storing an application's data, because remote-to-local memory transfers over a modern interconnection network are faster than traditional disk-to-memory transfers. We explore the possibility of using the remote memory as (i) a (faster-than-disk) backing store, (ii) an extension of main memory accessed using single (remote) memory references, and (iii) as a combination of both. We use execution driven simulation to investigate the performance impact the use of remote memory has on several real programs. We conclude that even for today's low throughput networks, using remote memory as a place for storing (some) of an application's data may result in significant performance improvements, which will continue to get higher as the disparity between disk transfer rates and network transfer rates continues to increase.","PeriodicalId":256311,"journal":{"name":"Proceedings of MASCOTS '96 - 4th International Workshop on Modeling, Analysis and Simulation of Computer and Telecommunication Systems","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1996-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126433006","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Two new algorithms to compute steady-state bounds for Markov models with slow forward and fast backward transitions","authors":"J. Carrasco, A. Calderón, J. Escribá","doi":"10.1109/MASCOT.1996.500999","DOIUrl":"https://doi.org/10.1109/MASCOT.1996.500999","url":null,"abstract":"Two new algorithms are proposed for the computation of bounds for the steady-state reward rate of irreducible finite Markov models with slow forward and fast backward transitions. The algorithms use detailed knowledge of the model in a subset of generated states G and partial information about the model in the non-generated portion U of the state space. U is assumed partitioned into subsets U/sub k/, 1/spl les/k/spl les/N with a \"nearest neighbor\" structure. The algorithms involve the solution of, respectively, |M|+2 and 4 linear systems of size |G|, where M is the set of values of k corresponding to the subsets U/sub k/ through which the model can jump from G to U. Previously proposed algorithms for the same type of models required the solution of |S| linear systems of size |G|+N, where S is the subset of G through which the model can enter C from U, to achieve the same bounds as our algorithms, or gave less tighter bounds if state cloning techniques were used to reduce the number of solved linear systems. An availability model with system state dependent repair rates is used to illustrate the application and performance of the algorithms.","PeriodicalId":256311,"journal":{"name":"Proceedings of MASCOTS '96 - 4th International Workshop on Modeling, Analysis and Simulation of Computer and Telecommunication Systems","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1996-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131747889","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Performance analysis of WDM networks with wavelength conversion","authors":"Jyh-cheng Chen, P. Dowd","doi":"10.1109/MASCOT.1996.501008","DOIUrl":"https://doi.org/10.1109/MASCOT.1996.501008","url":null,"abstract":"The paper evaluates a circuit-switched multi-hop WDM network with wavelength conversion by a semi-Markov based approach. The proposed technique provides accurate performance prediction while also preserving model flexibility. Whether the wavelength conversion is beneficial or not depends on many factors. The main concerns considered are the number of wavelengths, packet generation rate, processing latency, and propagation delay. The network architecture evaluated is a bidirectional wrap-around mesh. System performance is measured in terms of network throughput and average packet delay. The results quantify the magnitude of performance improvement achieved through wavelength conversion.","PeriodicalId":256311,"journal":{"name":"Proceedings of MASCOTS '96 - 4th International Workshop on Modeling, Analysis and Simulation of Computer and Telecommunication Systems","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1996-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133169099","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
J., K. Flanagan, B. Nelson, Greg Thompson, J. Archibald, Brigham Young
{"title":"The inaccuracy of trace-driven simulation using incomplete multiprogramming trace data","authors":"J., K. Flanagan, B. Nelson, Greg Thompson, J. Archibald, Brigham Young","doi":"10.1109/MASCOT.1996.500991","DOIUrl":"https://doi.org/10.1109/MASCOT.1996.500991","url":null,"abstract":"Trace-driven simulation is commonly used to predict the performance of computer systems. However, existing tracing techniques produce traces inadequate for some studies: they do not usually record operating system references, and they produce relatively short traces. The paper explores the impact of these trace distortions on the performance estimates of uniprocessor memory hierarchies using multiprogramming workloads. We used a hardware monitor to capture traces under a variety of workloads and operating systems. Our monitor captures every reference and can. Record arbitrarily long traces. We quantify memory hierarchy performance using traces of the SPEC SDM1.1 benchmark suite executing on an i486 CPU. To evaluate variations due to operating systems, we compare these results under both Mach 3.0 and UNIX Sys V R4. We conclude that for current uniprocessors, long but incomplete traces result in modest errors in estimated performance, but for proposed architectures with large delays to main memory, the errors can be significant.","PeriodicalId":256311,"journal":{"name":"Proceedings of MASCOTS '96 - 4th International Workshop on Modeling, Analysis and Simulation of Computer and Telecommunication Systems","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1996-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122236553","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Event-based study of the effect of execution environments on parallel program performance","authors":"S. Sarukkai, Jerry C. Yan","doi":"10.1109/MASCOT.1996.501027","DOIUrl":"https://doi.org/10.1109/MASCOT.1996.501027","url":null,"abstract":"In this paper we seek to demonstrate the importance of studying the effect of changes in execution environment parameters, on parallel applications executed on state-of-the-art multiprocessors. A comprehensive methodology for event-based analysis of program behavior is introduced. This methodology is used to study the performance significance of various system parameters such as processor speed, message-buffer size, buffer copy speed, network bandwidth, communication latency, interrupt overheads and other system parameters. With the help of a few CFD examples, we illustrate the use of our technique in determining suitable parameter values of the execution environment for three applications.","PeriodicalId":256311,"journal":{"name":"Proceedings of MASCOTS '96 - 4th International Workshop on Modeling, Analysis and Simulation of Computer and Telecommunication Systems","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1996-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116650846","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Performability management in distributed database systems: an adaptive concurrency control protocol","authors":"A. Tai, J. F. Meyer","doi":"10.1109/MASCOT.1996.501020","DOIUrl":"https://doi.org/10.1109/MASCOT.1996.501020","url":null,"abstract":"An adaptive concurrency control protocol is proposed for distributed database systems where data and resource contention have adverse effects on performability. The protocol dynamically switches between two existing concurrency control mechanisms, namely, an optimistic one called the \"exclusive writer with locking option\" and a pessimistic one referred to as \"primary site locking.\" Adaptation is based on partial knowledge about the system that is available locally at any database site such that no overhead (no additional inter-database site message exchange for collecting information) is incurred. The evaluation results indicate a potential for significant gains in performability from the use of adaptive concurrency control techniques.","PeriodicalId":256311,"journal":{"name":"Proceedings of MASCOTS '96 - 4th International Workshop on Modeling, Analysis and Simulation of Computer and Telecommunication Systems","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1996-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123670710","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Measurement and modeling of EARTH-MANNA multithreaded architecture","authors":"S. Nemawarkar, G. Gao","doi":"10.1109/MASCOT.1996.501002","DOIUrl":"https://doi.org/10.1109/MASCOT.1996.501002","url":null,"abstract":"We develop and apply an analytical model to analyze the performance of the EARTH-MANNA multithreaded multiprocessor system. The performance model is based on closed queuing networks. We develop heuristics to account for the realistic subsystem interactions (simultaneous resource possession) and multithreaded workload. Inputs to the analytical model are architectural parameters of the EARTH-MANNA system, and program workload parameters. Predictions of the performance model include processor utilization and network latency. We characterize the variation of these performance measures in terms of program workload and architecture characteristics. We validate the analytical model through runtime measurements from the actual program executions on the EARTH-MANNA system.","PeriodicalId":256311,"journal":{"name":"Proceedings of MASCOTS '96 - 4th International Workshop on Modeling, Analysis and Simulation of Computer and Telecommunication Systems","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1996-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128206464","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Building parallel simulations from serial simulators","authors":"P. Heidelberger, D. Nicol","doi":"10.1109/MASCOT.1996.500986","DOIUrl":"https://doi.org/10.1109/MASCOT.1996.500986","url":null,"abstract":"In order for parallel discrete event simulation (PDES) to have practical impact, PDES tools must hide the complexities of time synchronisation from users and provide capabilities approaching those of industrial quality serial simulators. Typically, PDES tools have been started from scratch thereby requiring a modeler to learn a new and unfamiliar modeling tool. This note describes a parallel simulation tool, the Utilitarian Parallel Simulator, U.P.S. that is distinguished from the above mentioned tools in that it is built on top of an existing, industrial quality serial simulator, CSIM. U.P.S. extends the standard CSIM modeling language by providing a simple and natural mechanism for submodels running on different processors to communicate with one another. The user need not be concerned with time synchronisation, save for specifying which algorithm to use, nor need the user make calls to a low level interprocessor communications library (e.g. a message-passing library such as MPI). U.P.S. was implemented without making any changes to the existing CSIM simulation engine. As such, it implements conservative time synchronisation algorithms. As CSIM has been ported to many uniprocessor platforms, and since U.P.S. is implemented using the MPI standard, U.P.S. is potentially portable to many multiple processor systems. It has been tested and run on two such platforms: the IBM SP2 and the Intel Paragon.","PeriodicalId":256311,"journal":{"name":"Proceedings of MASCOTS '96 - 4th International Workshop on Modeling, Analysis and Simulation of Computer and Telecommunication Systems","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1996-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131115432","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}