Speeding Up Distributed MapReduce Applications Using Hardware Accelerators

Abstract

Abstract-In an attempt to increase the performance/cost ratio, large compute clusters are becoming heterogeneous at multiple levels: from asymmetric processors, to different system architectures, operating systems and networks. Exploiting the intrinsic multi-level parallelism present in such a complex execution environment has become a challenging task using traditional parallel and distributed programming models. As a result, an increasing need for novel approaches to exploiting parallelism has arisen in these environments. MapReduce is a data-driven programming model originally proposed by Google back in 2004 as a flexible alternative to the existing models, specially devoted to hiding the complexity of both developing and running massively distributed applications in large compute clusters. In some recent works, the MapReduce model has been also used to exploit parallelism in other non-distributed environments, such as multi-cores, heterogeneous processors and GPUs. In this paper we introduce a novel approach for exploiting the heterogeneity of a Cell BE cluster linking an existing MapReduce runtime implementation for distributed clusters and one runtime to exploit the parallelism of the Cell BE nodes. The novel contribution of this work is the design and evaluation of a MapReduce execution environment that effectively exploits the parallelism existing at both the Cell BE cluster level and the heterogeneous processors level.

CellMT: A cooperative multithreading library for the Cell/B.E.

“…In the future we plan to further investigate the suitability and feasibility of our threading library for applications with non predictable memory access such as list ranking and other combinatorial algorithms that cannot even use double buffering techniques. We will also try to further reduce the current context switch overhead to improve the performance of our library for even a wider range of applications, as well as, to integrate the CellMT threading library with other runtime systems supporting novel programming models, such as CellSs [8] and MapReduce [9] [16], or future implementations of OpenCL [10].…”

“…This makes it very attractive for any existing application that wants to make the most of the Cell/BE without increasing the code complexity. In addition to ease the development of end user applications, this library can also be used to simplify the implementation of runtime system and specialized programming frameworks such as [8], [9], [6] and [10].…”

Performance Management of Accelerated MapReduce Workloads in Heterogeneous Clusters

“…Heterogeneous hardware (mixing generic processors with accelerator cores such as GPUs or the SPUs in the Cell/BE [7] processor) will be leveraged to improve both performance and energy consumption, making the best of each specific platform. For example, a MapReduce framework enabled to run on hybrid systems, as the one introduced in [8], has the potential to impact immensely upon the future of many fields such as financial analysis, healthcare and smart cities management. The MapReduce framework and the domain-specific languages built upon it, provide an easy and convenient way to develop massively distributed data analytics services that exploit all the computing power of these large scale facilities, while low level generic languages, such as OpenCL [9], will provide portability across different hardware platforms.…”

“…Such an approach requires an integrated management of next generation data centers that addresses two critical goals: meeting high level performance goals for data analytics services, and exploiting the capabilities of heterogeneous hardware. While both challenges have been addressed in the past by separate (see [3] and [8] for more details) their integration represents a completely new challenge. Such integration is addressed in this paper.…”

Research on Hybrid Distributed Computing System Based on Embedded System

“…MIT and the University of Manchester researchers improved MapReduce performance on multi-core hardware [16]. The literature [17] proposed MapReduce on Cell Broadband Engine performance optimization techniques, and researchers at the University of Wisconsin used Cell Sort algorithm, giving full play to hardware capabilities, which greatly improved the performance of sorting. MapReduce performance on multi-core hardware improvements also includes in literature [18].…”