Multi-dimensional Self-tuning in Transactional Memory
نویسنده
چکیده
The Transactional Memory (TM) paradigm promises to greatly simplify the development of concurrent applications. This led, over the years, to the creation of a plethora of TM implementations delivering wide ranges of performance across workloads. Yet, no universal TM implementation fits each and every workload. In fact, the best TM in a given workload can reveal to be disastrous for another one. This forces developers to face the complex task of tuning TM implementations, which significantly hampers the wide adoption of TMs. This thesis addresses the challenge of automatically identifying the best TM implementation for a given workload. The proposed system, ProteusTM, hides behind the TM interface a large library of implementations. Under the hood, it leverages an innovative, multi-dimensional online optimization scheme, combining two popular machine learning techniques: Collaborative Filtering and Bayesian Optimization. ProteusTM was extensively evaluated, obtaining average performance < 3% from optimal, and gains up to 100× over static alternatives.
منابع مشابه
A recommendation system approach to the tuning of Transactional Memory
Transactional Memory (TM) is a promising approach that simplifies parallel programming. However, in the broad spectrum of available TM implementations, there exists no one size fits all solution that can provide optimal performance across all possible workloads. This has motivated an intense research, over the last years, on the design of self-tuning solutions aimed at transparently adapting th...
متن کاملMulti-dimensional Self-tuning in Transactional Memory
T r ansactional Memory is a powerful abstraction that promises to drastically simplify parallel programming. However, the efficiency of existing TM implementations can be strongly affected by the characteristics of the workloads generated by TM applications. This has raised interest in designing self-tuning solutions that adapt specific building blocks of existing TM implementations. This disse...
متن کاملSelf-tuning the parallelism degree in Parallel-Nested Transactional Memory
In this document we present an overview of Transactional Memory models, as well as a comparison to traditional parallel frameworks. In this context, we follow with an analysis on the problem of self-tuning concurrency control in transactional memory, where different mechanisms are used to predict and adjust an application’s concurrency level with respect to available parallelism. We are particu...
متن کاملSelf-tuning the parallelism degree in Parallel-Nested Software Transactional Memory
Transactional Memory is a promising parallel computing paradigm, increasingly used nowadays, that allows easy parallelization of sequential programs and can expose a great level of parallelism. Many different approaches exist. One such approach is parallel-nested Transactional Memory, where transactions are allowed to spawn their own child transactions, effectively exposing additional levels of...
متن کاملSelf-tuning Intel Restricted Transactional Memory
The Transactional Memory (TM) paradigm aims at simplifying the development of concurrent applications by means of the familiar abstraction of atomic transaction. After a decade of intense research, hardware implementations of TM have recently entered the domain of mainstream computing thanks to Intel’s decision to integrate TM support, codenamed RTM (Reduced Transactional Memory), in their last...
متن کامل