Language and Virtual Machine Support for Efficient Fine-Grained Futures in Java

Abstract

In this work, we investigate the implementation of futures in Java J2SE v5.0. Java 5.0 provides an interface-based implementation of futures that enables users to encapsulate potentially asynchronous computation and to define their own execution engines for futures. Although this methodology decouples thread scheduling from application logic, for applications with fine-grained parallelism, this model imposes an undue burden on the average users and introduces significant performance overhead.To address these issues, we investigate the use of lazy futures and offer an alternative implementation to the Java 5.0 approach. In particular, we present a directive-based programming model for using futures in Java that uses annotations in Java 5.0 (as opposed to interfaces) and a lazy future implementation to significantly simplify programmer effort. Our directive-based future system employs novel compilation and runtime techniques that transparently and adaptively split and spawn futures for parallel execution. All such decisions are automatic and guided by dynamically determined future granularity and underlying resource availability. We empirically evaluate our future implementation using different Java Virtual Machine configurations and common Java benchmarks that implement fine-grained parallelism. We compare directive-based lazy futures with lazy and Java 5.0 futures and show that our approach is significantly more scalable.

Supporting exception handling for futures in Java

“…To simplify programmer effort and to improve performance and scalability of future-based applications in Java, in prior work [39], we propose a new implementation of futures in Java that we refer to as Directive-based Lazy Futures (DBLFutures).…”

“…We provide a complete description of the DBLFuture implementation in [39] and [40]. In summary, our DBLFuture system facilitates:…”

“…It was first introduced in Multilisp [16], and has been supported by many languages including Java J2SE 5.0 [23] and X10 [8]. In our prior work [39], we propose a new implementation of futures in Java that we refer to as Directive-based Lazy Futures, and DBLFutures for short. In the DBLFuture model, programmers use a future directive, denoted as a new Java annotation, @future, to specify potentially concurrent computations within a serial program, and leave the decisions of when and how to execute these computations to the Java virtual machine (JVM).…”

“…We begin with the existing Future APIs in Java 5.0, and then give an overview of directive-based lazy futures, a technique that we have developed as part of prior work [39].…”

Parallelism Granules Aggregation with the T-System

“…"Task inlining" [3], "lazy task creation" [4] and "leapfrogging" [5] has been devised almost two decades ago for Mul-T [3] and Multilisp [6]. In principle, all these techniques are applicable in context of any futures-based system, like it has been recently shown for Java-based system in [7]. A lot of work has been done on granularity control in Glasgow Parallel Haskell both in terms of CPU and memory resources [8,9].…”