Optimizing dataflow applications on heterogeneous environments

2014 IEEE 26th International Symposium on Computer Architecture and High Performance Computing

et al. 2014

Self Cite

High performance computing is experiencing a major paradigm shift with the introduction of accelerators, such as graphics processing units (GPUs) and Intel Xeon Phi (MIC). These processors have made available a tremendous computing power at low cost, and are transforming machines into hybrid systems equipped with CPUs and accelerators. Although these systems can deliver a very high peak performance, making full use of its resources in real-world applications is a complex problem. Most current applications deployed to these machines are still being executed in a single processor, leaving other devices underutilized. In this paper we explore a scenario in which applications are composed of hierarchical data flow tasks which are allocated to nodes of a distributed memory machine in coarse-grain, but each of them may be composed of several finer-grain tasks which can be allocated to different devices within the node. We propose and implement novel performance aware scheduling techniques that can be used to allocate tasks to devices. We evaluate our techniques using a pathology image analysis application used to investigate brain cancer morphology, and our experimental evaluation shows that the proposed scheduling strategies significantly outperforms other efficient scheduling techniques, such as Heterogeneous Earliest Finish Time - HEFT, in cooperative executions using CPUs, GPUs, and MICs. We also experimentally show that our strategies are less sensitive to inaccuracy in the scheduling input data and that the performance gains are maintained as the application scales.

Section: Related Workmentioning

confidence: 99%

Efficient Execution of Microscopy Image Analysis on CPU, GPU, and MIC Equipped Cluster Systems

Andrade

Ferreira

2014 IEEE 26th International Symposium on Computer Architecture and High Performance Computing

et al. 2014

Self Cite

“…There are several recent efforts on task scheduling for hybrid machines [17,19,24,26,9,27,15,28]. Most of the previous works deal with task mapping for applications in which operations attain similar speedups when executed on a GPU vs a CPU.…”

Section: Related Workmentioning

confidence: 99%

Haggis

Aji

Proceedings of the 3rd ACM SIGSPATIAL International Workshop on Analytics for Big Geospatial Data

Wang

2014

Self Cite

Spatial query processing involves complex multidimensional objects and compute intensive spatial operations, and therefore requires a high performance approach to meet the rapid data analytics requirements of modern spatial applications. Recently, MapReduce based spatial query systems have become a viable solution for many data intensive query tasks, and gained widespread adoption in both academia and industry. At the same time, GPUs have been successfully utilized in many applications that require high performance computation. Both approaches, GPU and MapReduce, have their own limitations and advantages, and have been separately utilized in spatial query processing tasks to boost application performance. However, it is unclear that how MapReduce and GPU, two vastly different parallelization techniques, can be fused together to effectively deal with the spatial big data challenges. In this paper, we explore such synergy of parallelization techniques for large scale spatial query processing. We extend Hadoop-GIS, a MapReduce based spatial query system, and provide GPU accelerated spatial query processing capability at the engine level. We evaluate the system on a real world dataset, and demonstrate that GPU accelerated system can gain considerable performance improvements. We also show how other factors such as partition granularity, task scheduling between CPU and GPU can impact the query performance.

“…Efficient execution of applications on distributed CPUGPU equipped platforms has been an objective of several projects [23], [24], [25], [26], [22], [29], [30]. Ravi et al [24], [26] proposes techniques for automatic translation of generalized reductions to CPU-GPU environments via compile-time techniques, which are coupled with runtime support to coordinate execution.…”

Section: Related Workmentioning

confidence: 99%

High-throughput Analysis of Large Microscopy Image Datasets on CPU-GPU Cluster Platforms

2013 IEEE 27th International Symposium on Parallel and Distributed Processing

Pan

Kurç

et al. 2013

Self Cite

Analysis of large pathology image datasets offers significant opportunities for the investigation of disease morphology, but the resource requirements of analysis pipelines limit the scale of such studies. Motivated by a brain cancer study, we propose and evaluate a parallel image analysis application pipeline for high throughput computation of large datasets of high resolution pathology tissue images on distributed CPU-GPU platforms. To achieve efficient execution on these hybrid systems, we have built runtime support that allows us to express the cancer image analysis application as a hierarchical data processing pipeline. The application is implemented as a coarse-grain pipeline of stages, where each stage may be further partitioned into another pipeline of fine-grain operations. The fine-grain operations are efficiently managed and scheduled for computation on CPUs and GPUs using performance aware scheduling techniques along with several optimizations, including architecture aware process placement, data locality conscious task assignment, data prefetching, and asynchronous data copy. These optimizations are employed to maximize the utilization of the aggregate computing power of CPUs and GPUs and minimize data copy overheads. Our experimental evaluation shows that the cooperative use of CPUs and GPUs achieves significant improvements on top of GPU-only versions (up to 1.6×) and that the execution of the application as a set of fine-grain operations provides more opportunities for runtime optimizations and attains better performance than coarser-grain, monolithic implementations used in other works. An implementation of the cancer image analysis pipeline using the runtime support was able to process an image dataset consisting of 36,848 4Kx4K-pixel image tiles (about 1.8TB uncompressed) in less than 4 minutes (150 tiles/second) on 100 nodes of a state-of-the-art hybrid cluster system.