Asymmetric multi-core processors (AMPs) with general-purpose and specialized cores packaged on the same chip, are emerging as a leading paradigm for high-end computing. A large body of exist- ing research explores the use of standalone AMPs in computationally challenging and data-intensive applications. AMPs are rapidly deployed as high-performance accelerators on clusters. In these set-tings, scheduling, communication and I/O are managed by general- purpose processors (GPPs), while computation is off-loaded to AMPs. Design space exploration for the configuration and software stack of hybrid clusters of AMPs and GPPs is an open problem. In this paper, we explore this design space in an implementation of the popular MapReduce programming model. Our contributions are: An exploration of various design alternatives for hybrid asymmetric clusters of AMPs and GPPs; the adoption of a streaming approach to supporting MapReduce computations on clusters with asymmet- ric components; and adaptive schedulers that take into account individual component capabilities in asymmetric clusters. Throughout our design, we remove I/O bottlenecks, using double-buffering and asynchronous I/O. We present an evaluation of the design choices through experiments on a real cluster with MapReduce workloads of varying degrees of computation intensity. We find that in a cluster with resource-constrained and well-provisioned AMP accelerators, a streaming approach achieves 50 . 5% and 73 . 1% bet- ter performance compared to the non-streaming approach, respectively, and scales almost linearly with increasing number of compute nodes. We also show that our dynamic scheduling mechanisms adapt effectively the parameters of the scheduling policies between applications with different computation density.

Resource management on accelerator based systems is complicated by the disjoint nature of the main CPU and accelerator, which involves separate memory hierarhcies, different degrees of parallelism, and relatively high cost of communicating between them. For applications with irregul parallelism, where work is dynamically created based on other computations, the accelerators may both consume and produce work. To maintain load balance, the accelerators hand work back to the CPU to be scheduled. In this paper we consider multiple approaches for such scheduling problems and use the Cell BE system to demonstrate the different schedulers and the trade-offs between them. Our evaluation is done with both microbenchmarks and two bioinformatics applications (PBPI and RAxML). Our baseline approach uses a standard Linux scheduler on the CPU, possibly with more than one process per CPU. We then consider the addition of cooperative scheduling to the Linux kernel and a user-level work-stealing approach. The two cooperative approaches are able to decrease SPE idle time, by 30% and 70%, respectively, relative to the baseline scheduler. In both cases we believe the changes required to application level codes, e.g., a program written with MPI processes that use accelerator based compute nodes, is reasonable, although the kernel level approach provides more generality and ease of implementation, but often less performance than work stealing approach.

On multiprocessors with explicitly managed memory hierarchies (EMM), software has the responsibility of moving data in and out of fast local memories. This task can be complex and error-prone even for expert programmers. Before we can allow compilers to handle this complexity for us, we must identify the abstractions that are general enough to allow us to write applications with reasonable effort, yet specific enough to exploit the vast on-chip memory bandwidth of EMM multi-processors. To this end, we compare two programming models against hand-tuned codes on the STI Cell, paying attention to programmability and performance. The first programming model, Sequoia, abstracts the memory hierarchy as private address spaces, each corresponding to a parallel task. The second, Cellgen, is a new framework which provides OpenMP-like semantics and the abstraction of a shared address space divided into private and shared data. We compare three applications programmed using these models against their hand-optimized counterparts in terms of abstractions, programming complexity, and performance.

On multiprocessors with explicitly managed memory hierarchies (EMM), software has the responsibility of moving data in and out of fast local memories. This task can be complex and error-prone even for expert programmers. Before we can allow compilers to handle this complexity for us, we must identify the abstractions that are general enough to allow us to write applications with reasonable effort, yet specific enough to exploit the vast on-chip memory bandwidth of EMM multi-processors. To this end, we compare two programming models against hand-tuned codes on the STI Cell, paying attention to programmability and performance. The first programming model, Sequoia, abstracts the memory hierarchy as private address spaces, each corresponding to a parallel task. The second, Cellgen, is a new framework which provides OpenMP-like semantics and the abstraction of a shared address space divided into private and shared data. We compare three applications programmed using these models against their hand-optimized counterparts in terms of abstractions, programming complexity, and performance.

Asymmetric multi-core processors (AMPs) with general-purpose and specialized cores packaged on the same chip, are emerging as a leading paradigm for high-end computing. A large body of existing research explores the use of standalone AMPs in computationally challenging and data-intensive applications. AMPs are rapidly deployed as high-performance accelerators on clusters. In these settings, scheduling, communication and I/O are managed by generalpurpose processors (GPPs), while computation is off-loaded to AMPs. Design space exploration for the configuration and software stack of hybrid clusters of AMPs and GPPs is an open problem. In this paper, we explore this design space in an implementation of the popular MapReduce programming model. Our contributions are: An exploration of various design alternatives for hybrid asymmetric clusters of AMPs and GPPs; the adoption of a streaming approach to supporting MapReduce computations on clusters with asymmetric components; and adaptive schedulers that take into account individual component capabilities in asymmetric clusters. Throughout our design, we remove I/O bottlenecks, using double-buffering and asynchronous I/O. We present an evaluation of the design choices through experiments on a real cluster with MapReduce workloads of varying degrees of computation intensity. We find that in a cluster with resource-constrained and well-provisioned AMP accelerators, a streaming approach achieves 50.5% and 73.1% better performance compared to the non-streaming approach, respectively, and scales almost linearly with increasing number of compute nodes.We also show that our dynamic scheduling mechanisms adapt effectively the parameters of the scheduling policies between applications with different computation density.

The use of asymmetric multi-core processors with on-chip computational accelerators is becoming common in a variety of environments ranging from scientific computing to enterprise applications. The focus of current research has been on making efficient use of individual systems, and porting applications to asymmetric processors. In this paper, we take the next step by investigating the use of multi-core-based systems, especially the popular Cell processor, in a cluster setting. We present CellMR, an efficient and scalable implementation of the MapReduce framework for asymmetric Cell-based clusters. The novelty of CellMR lies in its adoption of a streaming approach to supporting MapReduce, and its adaptive resource scheduling schemes: Instead of allocating workloads to the components once, CellMR slices the input into small work units and streams them to the asymmetric nodes for efficient processing. Moreover, CellMR removes I/O bottlenecks by design, using a number of techniques, such as double-buffering and asynchronous I/O, to maximize cluster performance. Our evaluation of CellMR using typical MapReduce applications shows that it achieves 50.5% better performance compared to the standard nonstreaming approach, introduces a very small overhead on the manager irrespective of application input size, scales almost linearly with increasing number of compute nodes (a speedup of 6.9 on average, when using eight nodes compared to a single node), and adapts effectively the parameters of its resource management policy between applications with varying computation density.