Online auction sites have unique workloads and user behavior characteristics that do not exist in other e-commerce sites. Earlier studies by the authors identified i) significant changes in the workload depending on time of day and ii) the presence of heavy-tailed distributions involving bidder and seller behavior. Other studies indicate that a small fraction or power users, among millions of registered users, contribute to the majority of the site's revenue. Poor quality of service to power users may imply in loss of business and cause significant loss of revenue to an auction site. This problem could be mitigated by dedicating resources in an exclusive manner to this small group of users in order to improve the quality of service they receive. However, this approach can lead to i) under utilization of these dedicated resources and ii) overloading of resources allocated to regular users when the load from power users is low. In this paper, we propose a scheme whereby resources are primarily dedicated to power users, but regular users can take advantage of these dedicated resources when spare capacity is available. This paper provides a business-oriented framework for dispatching requests to the various servers of an online auction site. Our approach uses a controller that can dynamically shift the load to different types of servers as the workload changes. Experimental evaluation showed, among other things, that the total number of bids processed using a dynamic controller to allocate resources can be improved under heavy load conditions compared to using a load balancing technique that does not differentiate among users.
{"title":"A business-oriented load dispatching framework for online auction sites","authors":"D. Menascé, V. Akula","doi":"10.1109/QEST.2007.28","DOIUrl":"https://doi.org/10.1109/QEST.2007.28","url":null,"abstract":"Online auction sites have unique workloads and user behavior characteristics that do not exist in other e-commerce sites. Earlier studies by the authors identified i) significant changes in the workload depending on time of day and ii) the presence of heavy-tailed distributions involving bidder and seller behavior. Other studies indicate that a small fraction or power users, among millions of registered users, contribute to the majority of the site's revenue. Poor quality of service to power users may imply in loss of business and cause significant loss of revenue to an auction site. This problem could be mitigated by dedicating resources in an exclusive manner to this small group of users in order to improve the quality of service they receive. However, this approach can lead to i) under utilization of these dedicated resources and ii) overloading of resources allocated to regular users when the load from power users is low. In this paper, we propose a scheme whereby resources are primarily dedicated to power users, but regular users can take advantage of these dedicated resources when spare capacity is available. This paper provides a business-oriented framework for dispatching requests to the various servers of an online auction site. Our approach uses a controller that can dynamically shift the load to different types of servers as the workload changes. Experimental evaluation showed, among other things, that the total number of bids processed using a dynamic controller to allocate resources can be improved under heavy load conditions compared to using a load balancing technique that does not differentiate among users.","PeriodicalId":249627,"journal":{"name":"Fourth International Conference on the Quantitative Evaluation of Systems (QEST 2007)","volume":"112 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131089306","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}
L. D. Alfaro, K. Chatterjee, M. Faella, Axel Legay
We present qualitative randomized CTL (QRCTL), a qualitative version of pCTL, for specifying properties of Markov decision processes (MDPs). QRCTL formulas can express the fact that certain temporal properties hold with probability 0 or 1, but they do not distinguish other probabilities values. We present a symbolic, polynomial time model-checking algorithm for QRCTL on MDPs. Then, we study the equivalence relation induced by QRCTL, called qualitative equivalence. We show that for finite alternating MDPs, where nondeterministic and probabilistic choice occur in different states, qualitative equivalence coincides with alternating bisimulation, and can thus be computed via efficient partition-refinement algorithms. Surprisingly, the result does not hold for non-alternating MDPs. Indeed, we show that no local partition refinement algorithm can compute qualitative equivalence on non-alternating MDPs. Finally, we consider QRCTL*, that is the "star extension" of QRCTL. We show that QRCTL and QRCTL* induce the same qualitative equivalence on alternating MDPs, while on non-alternating MDPs, the equivalence arising from QRCTL* can be strictly finer. We also provide a full characterization of the relation between qualitative equivalence, bisimulation, and alternating bisimulation, according to whether the MDPs are finite, and to whether their transition relations are finite-branching.
{"title":"Qualitative Logics and Equivalences for Probabilistic Systems","authors":"L. D. Alfaro, K. Chatterjee, M. Faella, Axel Legay","doi":"10.2168/LMCS-5(2:7)2009","DOIUrl":"https://doi.org/10.2168/LMCS-5(2:7)2009","url":null,"abstract":"We present qualitative randomized CTL (QRCTL), a qualitative version of pCTL, for specifying properties of Markov decision processes (MDPs). QRCTL formulas can express the fact that certain temporal properties hold with probability 0 or 1, but they do not distinguish other probabilities values. We present a symbolic, polynomial time model-checking algorithm for QRCTL on MDPs. Then, we study the equivalence relation induced by QRCTL, called qualitative equivalence. We show that for finite alternating MDPs, where nondeterministic and probabilistic choice occur in different states, qualitative equivalence coincides with alternating bisimulation, and can thus be computed via efficient partition-refinement algorithms. Surprisingly, the result does not hold for non-alternating MDPs. Indeed, we show that no local partition refinement algorithm can compute qualitative equivalence on non-alternating MDPs. Finally, we consider QRCTL*, that is the \"star extension\" of QRCTL. We show that QRCTL and QRCTL* induce the same qualitative equivalence on alternating MDPs, while on non-alternating MDPs, the equivalence arising from QRCTL* can be strictly finer. We also provide a full characterization of the relation between qualitative equivalence, bisimulation, and alternating bisimulation, according to whether the MDPs are finite, and to whether their transition relations are finite-branching.","PeriodicalId":249627,"journal":{"name":"Fourth International Conference on the Quantitative Evaluation of Systems (QEST 2007)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115434368","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}
M. Beccuti, D. Raiteri, G. Franceschinis, S. Haddad
The Markov decision process (MDP) (M.L. Puterman, 2005) formalism is widely used for modeling systems which exhibit both non deterministic and probabilistic behaviors (e.g. distributed systems, resource management systems, ...). Unfortunately, if the system is particularly complex then its modeling at the MDP level may be very hard; so in (M. Beccuti et al., 2007) a higher-level formalism called Markov decision well-formed net (MDWN) was proposed. The MDWN allows to describe the system in terms of its components and their interactions, while the MDP describes directly the state space and the state transitions. The MDWN model is more compact and readable: in particular, it is possible to define a complex non deterministic or probabilistic behavior as a composition of simpler non deterministic or probabilistic steps. In the MDWN formalism, the probabilistic behavior of the system is clearly distinct from the non deterministic one; actually they are designed as two separate Petri nets (PN): the probabilistic PN (Npr) and the non deterministic PN (Nnd).
{"title":"A framework to design and solve Markov Decision Well-formed Net models","authors":"M. Beccuti, D. Raiteri, G. Franceschinis, S. Haddad","doi":"10.1109/QEST.2007.32","DOIUrl":"https://doi.org/10.1109/QEST.2007.32","url":null,"abstract":"The Markov decision process (MDP) (M.L. Puterman, 2005) formalism is widely used for modeling systems which exhibit both non deterministic and probabilistic behaviors (e.g. distributed systems, resource management systems, ...). Unfortunately, if the system is particularly complex then its modeling at the MDP level may be very hard; so in (M. Beccuti et al., 2007) a higher-level formalism called Markov decision well-formed net (MDWN) was proposed. The MDWN allows to describe the system in terms of its components and their interactions, while the MDP describes directly the state space and the state transitions. The MDWN model is more compact and readable: in particular, it is possible to define a complex non deterministic or probabilistic behavior as a composition of simpler non deterministic or probabilistic steps. In the MDWN formalism, the probabilistic behavior of the system is clearly distinct from the non deterministic one; actually they are designed as two separate Petri nets (PN): the probabilistic PN (Npr) and the non deterministic PN (Nnd).","PeriodicalId":249627,"journal":{"name":"Fourth International Conference on the Quantitative Evaluation of Systems (QEST 2007)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121596259","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}
Probabilistic models are widely used to analyze embedded, networked, and more recently biological systems. Existing numerical analysis techniques are limited to finite- state models and suffer from the state explosion problem. As a consequence, the user often has to manually abstract the intended model to get a tractable one. To this end, we propose the predicate abstraction model checker PASS which automates this process. We leverage recent advances in automatic theorem proving to compute tractable finite-state models. Experiments show the feasibility of our approach. To the best of our knowledge, this is the first time that properties of probabilistic infinite-state models have been verified at this level of automation.
{"title":"Probabilistic Model Checking Modulo Theories","authors":"Björn Wachter, Lijun Zhang, H. Hermanns","doi":"10.1109/QEST.2007.10","DOIUrl":"https://doi.org/10.1109/QEST.2007.10","url":null,"abstract":"Probabilistic models are widely used to analyze embedded, networked, and more recently biological systems. Existing numerical analysis techniques are limited to finite- state models and suffer from the state explosion problem. As a consequence, the user often has to manually abstract the intended model to get a tractable one. To this end, we propose the predicate abstraction model checker PASS which automates this process. We leverage recent advances in automatic theorem proving to compute tractable finite-state models. Experiments show the feasibility of our approach. To the best of our knowledge, this is the first time that properties of probabilistic infinite-state models have been verified at this level of automation.","PeriodicalId":249627,"journal":{"name":"Fourth International Conference on the Quantitative Evaluation of Systems (QEST 2007)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122702891","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}
T. Courtney, Shravan Gaonkar, M. McQuinn, Eric Rozier, W. Sanders, P. Webster
Models of complex systems often contain model parameters for important rates, probabilities, and initial state values. By varying the parameter values, the system modeler can study the behavior of the system under a wide range of system and environmental assumptions. However, exhaustive exploration of the parameter space of a large model is computationally expensive. Design of experiments techniques provide information about the degree of sensitivity of output variables to various input parameters. Design of experiments makes it possible to find parameter values that optimize measured outputs of the system by running fewer experiments than required by less rigorous techniques. This paper describes the design of experiments techniques that have been integrated in the Mobius tool.
{"title":"Design of Experiments within the Mobius Modeling Environment","authors":"T. Courtney, Shravan Gaonkar, M. McQuinn, Eric Rozier, W. Sanders, P. Webster","doi":"10.1109/QEST.2007.36","DOIUrl":"https://doi.org/10.1109/QEST.2007.36","url":null,"abstract":"Models of complex systems often contain model parameters for important rates, probabilities, and initial state values. By varying the parameter values, the system modeler can study the behavior of the system under a wide range of system and environmental assumptions. However, exhaustive exploration of the parameter space of a large model is computationally expensive. Design of experiments techniques provide information about the degree of sensitivity of output variables to various input parameters. Design of experiments makes it possible to find parameter values that optimize measured outputs of the system by running fewer experiments than required by less rigorous techniques. This paper describes the design of experiments techniques that have been integrated in the Mobius tool.","PeriodicalId":249627,"journal":{"name":"Fourth International Conference on the Quantitative Evaluation of Systems (QEST 2007)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130906280","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. Barnat, L. Brim, I. Cerná, Milan Ceska, Jana Tumova
introduce a parallel model checker for checking Markov decision processes against linear time properties. The model checker extends the parallel model checker DiVinE and supports verification of qualitative properties.
{"title":"ProbDiVinE: A Parallel Qualitative LTL Model Checker","authors":"J. Barnat, L. Brim, I. Cerná, Milan Ceska, Jana Tumova","doi":"10.1109/QEST.2007.13","DOIUrl":"https://doi.org/10.1109/QEST.2007.13","url":null,"abstract":"introduce a parallel model checker for checking Markov decision processes against linear time properties. The model checker extends the parallel model checker DiVinE and supports verification of qualitative properties.","PeriodicalId":249627,"journal":{"name":"Fourth International Conference on the Quantitative Evaluation of Systems (QEST 2007)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131360825","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}
We have designed a tool to partition a Markov Chain and have used the censoring technique and strong stochastic comparison to obtain bounds on rewards and the first passage time. We present the main ideas of the method, the algorithms, the tool and some numerical results.
{"title":"CUT : Combining stochastic ordering and censoring to bound steady-state rewards and first passage time","authors":"J. Fourneau, N. Pekergin, S. Younès","doi":"10.1109/QEST.2007.35","DOIUrl":"https://doi.org/10.1109/QEST.2007.35","url":null,"abstract":"We have designed a tool to partition a Markov Chain and have used the censoring technique and strong stochastic comparison to obtain bounds on rewards and the first passage time. We present the main ideas of the method, the algorithms, the tool and some numerical results.","PeriodicalId":249627,"journal":{"name":"Fourth International Conference on the Quantitative Evaluation of Systems (QEST 2007)","volume":"86 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128693689","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}
C. U. Smith, Catalina M. Lladó, R. Puigjaner, L. Williams
XML-based interchange formats for performance models provide a mechanism whereby performance model information may be transferred among modeling tools. For example, the PMIF allows diverse tools to exchange queueing network model information. Formats have also been defined for the interchange of LQN, UML, Petri Nets, and others. These formats specify the model and a set of parameters for one run. For model studies, however, it is useful to be able to specify multiple runs, or experiments, for the model. This paper presents an XML interchange schema extension for defining a set of model runs and the output desired from them. It has the expressive power to specify iterations, alternations, assignments of values, actions based on model results and more. Examples illustrate how the experiment interchange extension can be used with a wide variety of performance modeling paradigms. A prototype proves the concept.
{"title":"Interchange Formats for Performance Models: Experimentation and Ouput","authors":"C. U. Smith, Catalina M. Lladó, R. Puigjaner, L. Williams","doi":"10.1109/QEST.2007.24","DOIUrl":"https://doi.org/10.1109/QEST.2007.24","url":null,"abstract":"XML-based interchange formats for performance models provide a mechanism whereby performance model information may be transferred among modeling tools. For example, the PMIF allows diverse tools to exchange queueing network model information. Formats have also been defined for the interchange of LQN, UML, Petri Nets, and others. These formats specify the model and a set of parameters for one run. For model studies, however, it is useful to be able to specify multiple runs, or experiments, for the model. This paper presents an XML interchange schema extension for defining a set of model runs and the output desired from them. It has the expressive power to specify iterations, alternations, assignments of values, actions based on model results and more. Examples illustrate how the experiment interchange extension can be used with a wide variety of performance modeling paradigms. A prototype proves the concept.","PeriodicalId":249627,"journal":{"name":"Fourth International Conference on the Quantitative Evaluation of Systems (QEST 2007)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130333017","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}
Previous studies have shown that buffering packets in DRAM is a performance bottleneck. In order to understand the impediments in accessing the DRAM, we developed a detailed Petri net model of IP forwarding application on IXP2400 that models the different levels of the memory hierarchy. The cell based interface used to receive and transmit packets in a network processor, leads to some small size DRAM accesses. Such narrow accesses to the DRAM expose the bank access latency, reducing the bandwidth that can be realized. With real traces up to 30% of the accesses are smaller than the cell size, resulting in 7.7% reduction in DRAM bandwidth. To overcome this problem, we propose buffering these small chunks of data in the on chip scratchpad memory. This scheme also exploits greater degree of parallelism between different levels of the memory hierarchy. Using real traces from the internet, we show that the transmit rate can be improved by an average of 21% over the base scheme without the use of additional hardware. Further, the impact of different traffic patterns on the network processor resources is studied. Under real traffic conditions, we show that the data bus which connects the off-chip packet buffer to the micro- engines, is the obstacle in achieving higher throughput.
{"title":"A Petri Net Model for Evaluating Packet Buffering Strategies in a Network Processor","authors":"C. GirishB., R. Govindarajan","doi":"10.1109/QEST.2007.29","DOIUrl":"https://doi.org/10.1109/QEST.2007.29","url":null,"abstract":"Previous studies have shown that buffering packets in DRAM is a performance bottleneck. In order to understand the impediments in accessing the DRAM, we developed a detailed Petri net model of IP forwarding application on IXP2400 that models the different levels of the memory hierarchy. The cell based interface used to receive and transmit packets in a network processor, leads to some small size DRAM accesses. Such narrow accesses to the DRAM expose the bank access latency, reducing the bandwidth that can be realized. With real traces up to 30% of the accesses are smaller than the cell size, resulting in 7.7% reduction in DRAM bandwidth. To overcome this problem, we propose buffering these small chunks of data in the on chip scratchpad memory. This scheme also exploits greater degree of parallelism between different levels of the memory hierarchy. Using real traces from the internet, we show that the transmit rate can be improved by an average of 21% over the base scheme without the use of additional hardware. Further, the impact of different traffic patterns on the network processor resources is studied. Under real traffic conditions, we show that the data bus which connects the off-chip packet buffer to the micro- engines, is the obstacle in achieving higher throughput.","PeriodicalId":249627,"journal":{"name":"Fourth International Conference on the Quantitative Evaluation of Systems (QEST 2007)","volume":"196 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126558289","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}
The paper introduces symbolic bisimulations for a simple probabilistic pi-calculus to overcome the infinite branching problem that still exists in checking ground bisimulations between probabilistic systems. Especially the definition of weak (symbolic) bisimulation does not rely on the random capability of adversaries and suggests a solution to the open problem on the axiomatization for weak bisimulation in the case of unguarded recursion. Furthermore, we present an efficient characterization of symbolic bisimulations for the calculus, which allows the "on-the-fly" instantiation of bound names and dynamic construction of equivalence relations for quantitative evaluation. This directly results in a local decision algorithm that can explore just a minimal portion of the state spaces of the probabilistic processes in question.
{"title":"Symbolic Bisimulations for Probabilistic Systems","authors":"Peng Wu, C. Palamidessi, Huimin Lin","doi":"10.1109/QEST.2007.11","DOIUrl":"https://doi.org/10.1109/QEST.2007.11","url":null,"abstract":"The paper introduces symbolic bisimulations for a simple probabilistic pi-calculus to overcome the infinite branching problem that still exists in checking ground bisimulations between probabilistic systems. Especially the definition of weak (symbolic) bisimulation does not rely on the random capability of adversaries and suggests a solution to the open problem on the axiomatization for weak bisimulation in the case of unguarded recursion. Furthermore, we present an efficient characterization of symbolic bisimulations for the calculus, which allows the \"on-the-fly\" instantiation of bound names and dynamic construction of equivalence relations for quantitative evaluation. This directly results in a local decision algorithm that can explore just a minimal portion of the state spaces of the probabilistic processes in question.","PeriodicalId":249627,"journal":{"name":"Fourth International Conference on the Quantitative Evaluation of Systems (QEST 2007)","volume":"127 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123449551","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}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：