Analytical/ML Mixed Approach for Concurrency Regulation in Software Transactional Memory

Abstract

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

An Analytical Model of Hardware Transactional Memory

“…A number of analytical models of concurrency control for database management systems have been proposed in the literature [35], [1], [37], [9]. More recently, several analytical models have been proposed for the concurrency control algorithms adopted by software implementations of TM [38], [21], [8], [14], [12], [32], [13]. The key difference with respect to these approaches is that in our model we consider peculiar characteristics of the concurrency control of HTM, including the co-existence of optimistic techniques (i.e., speculative execution of parallel transactions) and of a sequential/pessimistic fallback path.…”

Self-tuning in Distributed Transactional Memory

“…The problem of self-tuning TM has also been largely explored in literature, as TM and DTM, unsurprisingly, exhibit similar trade-offs, e.g., the workload characteristics can strongly affect the performance of the concurrency control algorithm, as well as the optimal MPL. Examples of self-tuning solutions that dynamically adjust these TM mechanisms/parameters can be found in [56,57,58,59].…”

“…Then, the ML is retrained over time in order to incorporate the knowledge coming from samples collected from the operational system [59,55].…”

Hybrid Machine Learning/Analytical Models for Performance Prediction

“…Gray box modeling [5,6,4,8,7,10,14,9,3] has emerged in the last years as an attempt to achieve the best of the AM and ML world. It relies on exploiting both methodologies, in order to compensate the weaknesses of the one with the strengths of the other.…”