Search

Software Transactional Memory (STM) stands as powerful concurrent programming paradigm, enabling atomicity, and isolation while accessing shared data. On the downside, STM may suffer from performance degradation due to excessive conflicts among concurrent transactions, which cause waste of CPU-cycles and energy because of transaction aborts. An approach to cope with this issue consists of putting in place smart scheduling strategies which temporarily suspend the execution of some transaction in order to reduce the transaction conflict rate. In this article, we present an adaptive model-based transaction scheduling technique relying on a Markov Chain-based performance model of STM systems. Our scheduling technique is adaptive in a twofold sense: (i) It controls the execution of transactions depending on throughput predictions by the model as a function of the current system state. (ii) It re-tunes on-line the Markov Chain-based model to adapt it—and the outcoming transaction scheduling decisions—to dynamic variations of the workload. We have been able to achieve the latter target thanks to the fact that our performance model is extremely lightweight. In fact, to be recomputed, it requires a reduced set of input parameters, whose values can be estimated via a few on-line samples related to the current workload dynamics. We also present a scheduler that implements our adaptive technique, which we integrated within the open source TinySTM package. Further, we report the results of an experimental study based on the STAMP benchmark suite, which has been aimed at assessing both the accuracy of our performance model in predicting the actual system throughput and the advantages of the adaptive scheduling policy over literature techniques.

In modern data centers, energy usage represents one of the major factors affecting operational costs. Power capping is a technique that limits the power consumption of individual systems, which allows reducing the overall power demand at both cluster and data center levels. However, literature power capping approaches do not fit well the nature of important applications based on first-class multi-thread technology. For these applications performance may not grow linearly as a function of the thread-level parallelism because of the need for thread synchronization while accessing shared resources, such as shared data. In this paper we consider the problem of maximizing the application performance under a power cap by dynamically tuning the thread-level parallelism and the power state of the CPU-cores. Based on experimental observations, we design an adaptive technique that selects in linear time the optimal combination of thread-level parallelism and CPU-core power state for the specific workload profile of the multi-threaded application. We evaluate our proposal by relying on different benchmarks, configured to use different thread synchronization methods, and compare its effectiveness to different state-of-the-art techniques., 11 pages, 18 figures

Transactional Memory (TM) stands as a powerful paradigm for manipulating shared data in concurrent applications. It avoids the drawbacks of coarse grain locking schemes, namely the potentially excessive limitation of concurrency, while jointly providing support for synchronization transparency to the programmers, which is achieved by embedding code-blocks accessing shared data within transactions. On the downside, excessive transaction aborts may arise in scenarios with non-negligible volumes of conflicting data accesses, which might significantly impair performance. TM needs therefore to resort to methods enabling applications to run with the maximum degree of transaction concurrency that still avoids thrashing. In this article, we focus on Software TM (STM) implementations and present a machine learning-based approach that enables the dynamic selection of the best suited number of threads to be kept alive along specific phases of the execution of STM applications, depending on (variations of) the shared data access pattern. Two key contributions are provided with our approach: (i) the identification of the well suited set of features allowing the instantiation of a reliable neural network-based performance model and (ii) the introduction of mechanisms enabling the reduction of the run-time overhead for sampling these features. We integrated a real implementation of our machine learning-based thread-parallelism regulation approach within the TinySTM open source package and present experimental data, based on the STAMP benchmark suite, which show the effectiveness of the presented thread-parallelism regulation policy in optimizing transaction throughput.

Software Transactional Memory (STM) may suffer from performance degradation due to excessive conflicts among concurrent transactions. An approach to cope with this issue consists in putting in place smart scheduling policies which temporarily suspend the execution of some transaction in order to reduce the actual conflict rate. In this paper, we present an adaptive transaction scheduling policy relying on a Markov Chain-based model of STM systems. The policy is adaptive in a twofold sense: (i) it schedules transactions depending on throughput predictions by the model as a function of the current system state; (ii) its underlying Markov Chain-based model is periodically re-instantiated at run-time to adapt it to dynamic variations of the workload. We also present an implementation of our adaptive transaction scheduler which has been integrated within the open source TinySTM package. The accuracy of our performance model in predicting the system throughput and the advantages of the adaptive scheduling policy over state-of-the-art approaches have been assessed via an experimental study based on the STAMP benchmark suite.

Synchronization transparency offered by Software Transactional Memory (STM) must not come at the expense of run-time efficiency, thus demanding from the STM-designer the inclusion of mechanisms properly oriented to performance and other quality indexes. Particularly, one core issue to cope with in STM is related to exploiting parallelism while also avoiding thrashing phenomena due to excessive transaction rollbacks, caused by excessively high levels of contention on logical resources, namely concurrently accessed data portions. A means to address run-time efficiency consists in dynamically determining the best-suited level of concurrency (number of threads) to be employed for running the application (or specific application phases) on top of the STM layer. For too low levels of concurrency, parallelism can be hampered. Conversely, over-dimensioning the concurrency level may give rise to the aforementioned thrashing phenomena caused by excessive data contention—an aspect which has reflections also on the side of reduced energy-efficiency. In this chapter we overview a set of recent techniques aimed at building “application-specific” performance models that can be exploited to dynamically tune the level of concurrency to the best-suited value. Although they share some base concepts while modeling the system performance vs the degree of concurrency, these techniques rely on disparate methods, such as machine learning or analytic methods (or combinations of the two), and achieve different tradeoffs in terms of the relation between the precision of the performance model and the latency for model instantiation. Implications of the different tradeoffs in real-life scenarios are also discussed.

In this paper, we present a novel framework for supporting the management and optimization of application subject to software anomalies and deployed on large scale cloud architectures, composed of different geographically distributed cloud regions. The framework uses machine learning models for predicting failures caused by accumulation of anomalies. It introduces a novel workload balancing approach and a proactive system scale up/scale down technique. We developed a prototype of the framework and present some experiments for validating the applicability of the proposed approaches.

A core problem affecting distributed data management systems relates to deciding the optimal system configuration in terms of, e.g., computing resources to be allocated. In the cloud computing era, this issue has become a compelling one, since cloud technologies enable elastic resource provisioning, based on dynamic ac- quisition and release of resources according to the "pay-per-use" pricing model. To take advantage of this resource management model, methods enabling the estima- tion of the minimum amount of resources that are required to sustain the application workload, while guaranteeing adequate system performance and availability (as es- tablished, e.g., in a Service Level Agreement - SLA), would be highly desirable. In this chapter, an overview of such a kind of methods is provided, particularly focusing on black-box system modeling approaches based on machine learning. Also, a case study where the automatic tuning of the configuration of a distributed in-memory data platform, deployed on top of a cloud infrastructure, is actuated via neural net- works is presented. In this case study, a controller, which exploits a neural-network based performance predictor, has been integrated with a mainstream open-source data platform and has been tested on top of the Amazon EC2 cloud platform.

In this article we exploit a combination of analytical and Machine Learning (ML) techniques in order to build a performance model allowing to dynamically tune the level of concurrency of applications based on Software Transactional Memory (STM). Our mixed approach has the advantage of reducing the training time of pure machine learning methods, and avoiding approximation errors typically affecting pure analytical approaches. Hence it allows very fast construction of highly reliable performance models, which can be promptly and effectively exploited for optimizing actual application runs. We also present a real implementation of a concurrency regulation architecture, based on the mixed modeling approach, which has been integrated with the open source Tiny STM package, together with experimental data related to runs of applications taken from the STAMP benchmark suite demonstrating the effectiveness of our proposal. © 2014 IEEE.

In this paper we explore machine-learning approaches for dynamically selecting the well suited amount of concurrent threads in applications relying on Software Transactional Memory (STM). Specifically, we present a solution that dynamically shrinks or enlarges the set of input features to be exploited by the machine-learner. This allows for tuning the concurrency level while also minimizing the overhead for input-features sampling, given that the cardinality of the input-feature set is always tuned to the minimum value that still guarantees reliability of workload characterization. We also present a fully heedged implementation of our proposal within the TinySTM open source framework, and provide the results of an experimental study relying on the STAMP benchmark suite, which show significant reduction of the response time with respect to proposals based on static feature selection. © 2014 IEEE. In this paper we explore machine-learning approaches for dynamically selecting the well suited amount of concurrent threads in applications relying on Software Transactional Memory (STM). Specifically, we present a solution that dynamically shrinks or enlarges the set of input features to be exploited by the machine-learner. This allows for tuning the concurrency level while also minimizing the overhead for input-features sampling, given that the cardinality of the input feature set is always tuned to the minimum value that still guarantees reliability of workload characterization. We also present a fully fledged implementation of our proposal within the TinySTM open source framework, and provide the results of an experimental study relying on the STAMP benchmark suite, which show significant reduction of the response time with respect to proposals based on static feature selection.

Elastic architectures and the "pay-as-you-go" resource pricing model offered by many cloud infrastructure providers may seem the right choice for companies dealing with data centric applications characterized by high variable workload. In such a context, in-memory transactional data grids have demonstrated to be particularly suited for exploiting advantages provided by elastic computing platforms, mainly thanks to their ability to be dynamically (re-)sized and tuned. Anyway, when specific QoS requirements have to be met, this kind of architectures have revealed to be complex to be managed by humans. Particularly, their management is a very complex task without the stand of mechanisms supporting run-time automatic sizing/tuning of the data platform and the underlying (virtual) hardware resources provided by the cloud. In this paper, we present a neural network-based architecture where the system is constantly and automatically re-configured, particularly in terms of computing resources, in order to achieve transaction class-based QoS while minimizing costs of the infrastructure. We also present some results showing the effectiveness of our architecture, which has been evaluated on top of Future Grid IaaS Cloud using Red Hat Infinispan in-memory data grid and the TPC-C benchmark.

Software Transactional Memory (STM) is recognized as an effective programming paradigm for concurrent applications. On the other hand, a core problem to cope with in STM deals with (dynamically) regulating the degree of concurrency, in order to deliver optimal performance. We address this problem by proposing a self-regulation approach of the concurrency level, which relies on a parametric analytical performance model aimed at predicting the scalability of the STM application as a function of the actual workload profile. The regulation scheme allows achieving optimal performance during the whole lifetime of the application via dynamic change of the number of concurrent threads according to the predictions by the model. The latter is customized for a specific application/platform through regression analysis, which is based on a lightweight sampling phase. We also present a real implementation of the model-based concurrency self-regulation architecture integrated within the open source TinySTM framework, and an experimental study based on standard STM benchmark applications. © 2013 IEEE. Software Transactional Memory (STM) is recognized as an effective programming paradigm for concurrent applications. On the other hand, a core problem to cope with in STM deals with (dynamically) regulating the degree of concurrency, in order to deliver optimal performance. We address this problem by proposing a self-regulation approach of the concurrency level, which relies on a parametric analytical performance model aimed at predicting the scalability of the STM application as a function of the actual workload profile. The regulation scheme allows achieving optimal performance during the whole lifetime of the application via dynamic change of the number of concurrent threads according to the predictions by the model. The latter is customized for a specific application/platform through regression analysis, which is based on a lightweight sampling phase. We also present a real implementation of the model-based concurrency self-regulation architecture integrated within the open source T

Cloud computing represents a cost-effective paradigm to deploy a wide class of large-scale distributed applications, for which the pay-per-use model combined with automatic resource provisioning promise to reduce the cost of dependability and scalability. However, a key challenge to be addressed to materialize the advantages promised by Cloud computing is the design of effective auto-scaling and self-tuning mechanisms capable of ensuring pre-determined QoS levels at minimum cost in face of changing workload conditions. This is one of the keys goals that are being pursued by the Cloud-TM project, a recent EU project that is developing a novel, self-optimizing transactional data platform for the cloud. In this paper we present the key design choices underlying the development of Cloud-TM's Workload Analyzer (WA), a crucial component of the Cloud-TM platform that is change of three key functionalities: aggregating, filtering and correlating the streams of statistical data gathered from the various nodes of the Cloud-TM platform, building detailed workload profiles of applications deployed on the Cloud-TM platform, characterizing their present and future demands in terms of both logical (i.e. data) and physical (e.g. hardware-related) resources, triggering alerts in presence of violations (or risks of future violations) of pre-determined SLAs. © 2012 IEEE.

One of the problems of Software-Transactional-Memory (STM) systems is the performance degradation that can be experienced when applications run with a non-optimal concurrency level, namely number of concurrent threads. When this level is too high a loss of performance may occur due to excessive data contention and consequent transaction aborts. Conversely, if concurrency is too low, the performance may be penalized due to limitation of both parallelism and exploitation of available resources. In this paper we propose a machine-learning based approach which enables STM systems to predict their performance as a function of the number of concurrent threads in order to dynamically select the optimal concurrency level during the whole lifetime of the application. In our approach, the STM is coupled with a neural network and an on-line control algorithm that activates or deactivates application threads in order to maximize performance via the selection of the most adequate concurrency level, as a function of the current data access profile. A real implementation of our proposal within the TinySTM open-source package and an experimental study relying on the STAMP benchmark suite are also presented. The experimental data confirm how our self-adjusting concurrency scheme constantly provides optimal performance, thus avoiding performance loss phases caused by non-suited selection of the amount of concurrent threads and associated with the above depicted phenomena. © 2012 IEEE.

We present an analytical performance modeling approach for concurrency control algorithms in the context of Software Transactional Memories (STMs). We consider a realistic execution pattern where each thread alternates the execution of transactional and non-transactional code portions. Our model captures dynamics related to the execution of both (i) transactional read/write memory accesses and (ii) non-transactional operations, even when they occur within transactional contexts. We rely on a detailed approach explicitly capturing key parameters, such as the execution cost of transactional and non-transactional operations, as well as the cost of begin, commit and abort operations. The proposed modeling methodology is general and extensible, lending itself to be easily specialized to capture the behavior of different STM concurrency control algorithms. In this work we specialize it to model the performance of Commit-Time-Locking algorithms, which are currently used by several STM systems. The presented analytical model has been validated against simulation results based on workload profiles derived by tracing applications proper of the STAMP benchmark suite, running on top of the TL2 transactional memory layer. (C) 2011 Elsevier B.V. All rights reserved.

Nowadays the 2-Phase-Locking (2PL) concurrency control algorithm still plays a core rule in the construction of transactional systems (e.g. database systems and transactional memories). Hence, any technique allowing accurate analysis and prediction of the performance of 2PL based systems can be of wide interest and applicability. In this article we present an accurate analytical model of 2PL concurrency control, which overcomes several limitations of preexisting analytical results. In particular our model captures relevant features of realistic data access patterns, by taking into account access distributions that depend on transactions' execution phases. Also, our model provides significantly more accurate performance predictions in heavy contention scenarios, where the number of transactions enqueued due to conflicting lock requests is expected to be non-minimal. The accuracy of our model has been verified against simulation results based on both synthetic data access patterns and patterns derived from the TPC-C benchmark. © 2009 ACM.

In this article we present a performance model for Multi-Version Concurrency Control (MVCC). This type of concurrency control is currently very popular among mainstream commercial and open source database systems thanks to its ability to well cope with read intensive workloads, as in the case of transaction profiles proper of Web applications. To build the model we had to tackle the intrinsic higher complexity of MVCC when compared to traditional concurrency control mechanisms (i.e. 2-Phase-Locking and optimistic ones), such as the joint use of locks and aborts to resolve direct conflicts among write accesses to the same data item, and the management of multiple data versions. We validate our analytical model via an extensive simulation study, considering both uniform and skewed data accesses, as well as differentiated transaction profiles. To the best of our knowledge, the present study provides the first analytical model of MVCC. © 2008 IEEE.