Markov Chain-Based Adaptive Scheduling in Software Transactional Memory

Abstract

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

Optimizing Job Coscheduling by Adaptive Deadlock-Free Scheduler

“…This increase causes the average number of cores that can be allocated by a user-level scheduler to decrease; i.e., there is a decrease in the exploitable parallelism that a node can provide for each user-level scheduler. A similar issue has been reported and studied in Software Transactional Memory (STM) [26,33,34]. The performance of a transactional system can be strongly affected by the conflict rate of transactions running concurrently, because a high degree of transactional concurrency may lead to an unacceptably high rate of aborts.…”

“…Our proposed heuristic-based approach is hard to guarantee convergence to the optimal solution for relieving performance degradation of coscheduling. We will explore and implement different approaches to optimize the scheduling phase of our proposed scheduler in the future study, such as the Markov Chain method proposed by previous study [26] and machine learning-based approach proposed by study [35].…”

“…Recently, it is considered as a new parallel programming mechanism in the transaction memory community [23,24]. Conventional programming models that apply lock-based concurrency control are known for their conceptual difficulty [25,26]. In particular, any programming errors that are produced by programmers can easily result in local or distributed deadlock.…”

“…The performance of a transactional system can be strongly affected by the conflict rate of transactions running concurrently, because a high degree of transactional concurrency may lead to an unacceptably high rate of aborts. The resolution of the above issues falls into two categories: performance model-based approaches [26,[35][36][37] and heuristic-based schemes [34,[38][39][40], which are summarized in Table 1.…”

An autonomic‐computing approach on mapping threads to multi‐cores for software transactional memory

“…These works either require offline training procedures to obtain an initial form of a function for parallelism prediction or demand to increment/decrement the thread number progressively in order to search its optimal value. A notable exception in the work of Sanzo et al 45 42 the parallelism degree is incremented cubically and later regulated linearly. During decrement, linear decrements are followed by multiplicative decrements.…”