Running multiple instances of the MapReduce framework concurrently in a multicluster system or datacenter enables data, failure, and version isolation, which is attractive for many organizations. It may also provide some form of performance isolation, but in order to achieve this in the face of time-varying workloads submitted to the MapReduce instances, a mechanism for dynamic resource (re-)allocations to those instances is required. In this paper, we present such a mechanism called Fawkes that attempts to balance the allocations to MapReduce instances so that they experience similar service levels. Fawkes proposes a new abstraction for deploying MapReduce instances on physical resources, the MR-cluster, which represents a set of resources that can grow and shrink, and that has a core on which MapReduce is installed with the usual data locality assumptions but that relaxes those assumptions for nodes outside the core. Fawkes dynamically grows and shrinks the active MR-clusters based on a family of weighting policies with weights derived from monitoring their operation.

MapReduce, which is the de facto programming model for large-scale distributed data processing, and its most popular implementation Hadoop have enjoyed widespread adoption in industry during the past few years. Unfortunately, from a performance point of view getting the most out of Hadoop is still a big challenge due to the large number of configuration parameters. Currently these parameters are tuned manually by trial and error, which is ineffective due to the large parameter space and the complex interactions among the parameters. Even worse, the parameters have to be re-tuned for different MapReduce applications and clusters. To make the parameter tuning process more effective, in this paper we explore machine learning-based performance models that we use to auto-tune the configuration parameters. To this end, we first evaluate several machine learning models with diverse MapReduce applications and cluster configurations, and we show that support vector regression model (SVR) has good accuracy and is also computationally efficient. We further assess our auto-tuning approach, which uses the SVR performance model, against the Starfish auto tuner, which uses a cost-based performance model. Our findings reveal that our auto-tuning approach can provide comparable or in some cases better performance improvements than Starfish with a smaller number of parameters. Finally, we propose and discuss a complete and practical end-to-end auto-tuning flow that combines our machine learning-based performance models with smart search algorithms for the effective training of the models and the effective exploration of the parameter space.

Cloud computing is an emerging infrastructure paradigm that promises to eliminate the need for companies to maintain expensive computing hardware. Through the use of virtualization and resource time-sharing, clouds address with a single set of physical resources a large user base with diverse needs. Thus, clouds have the potential to provide their owners the benefits of an economy of scale and, at the same time, become an alternative for both the industry and the scientific community to self-owned clusters, grids, and parallel production environments. For this potential to become reality, the first generation of commercial clouds need to be proven to be dependable. In this work we analyze the dependability of cloud services. Towards this end, we analyze long-term performance traces from Amazon Web Services and Google App Engine, currently two of the largest commercial clouds in production. We find that the performance of about half of the cloud services we investigate exhibits yearly and daily patterns, but also that most services have periods of especially stable performance. Last, through trace-based simulation we assess the impact of the variability observed for the studied cloud services on three large-scale applications, job execution in scientific computing, virtual goods trading in social networks, and state management in social gaming. We show that the impact of performance variability depends on the application, and give evidence that performance variability can be an important factor in cloud provider selection.

Interactive perception applications, such as gesture recognition and vision-based user interfaces, process high-data rate streams with compute intensive computer vision and machine learning algorithms. These applications can be represented as data flow graphs comprising several processing stages. Such applications require low latency to be interactive so that the results are immediately available to the user. To achieve low latency, we exploit the inherent coarse grained task and data parallelism of these applications by running them on clusters of machines. This paper addresses an important problem that arises: how to place the stages of these applications on machines to minimize the latency, and in particular, how to adjust an existing schedule in response to changes in the operating conditions (perturbations) while minimizing the disruption in the existing placement (churn). To this end, we propose four incremental placement heuristics which use the HEFT scheduling algorithm as their primary building block. Through simulations and experiments on a real implementation, using diverse workloads and a range of perturbation scenarios, we demonstrate that dynamic adjustment of the schedule can improve latency by as much as 36%, while producing little churn.

Cloud computing is an emerging commercial infrastructure paradigm that promises to eliminate the need for maintaining expensive computing facilities by companies and institutes alike. Through the use of virtualization and resource time sharing, clouds serve with a single set of physical resources a large user base with different needs. Thus, clouds have the potential to provide to their owners the benefits of an economy of scale and, at the same time, become an alternative for scientists to clusters, grids, and parallel production environments. However, the current commercial clouds have been built to support web and small database workloads, which are very different from typical scientific computing workloads. Moreover, the use of virtualization and resource time sharing may introduce significant performance penalties for the demanding scientific computing workloads. In this work, we analyze the performance of cloud computing services for scientific computing workloads. We quantify the presence in real scientific computing workloads of Many-Task Computing (MTC) users, that is, of users who employ loosely coupled applications comprising many tasks to achieve their scientific goals. Then, we perform an empirical evaluation of the performance of four commercial cloud computing services including Amazon EC2, which is currently the largest commercial cloud. Last, we compare through trace-based simulation the performance characteristics and cost models of clouds and other scientific computing platforms, for general and MTC-based scientific computing workloads. Our results indicate that the current clouds need an order of magnitude in performance improvement to be useful to the scientific community, and show which improvements should be considered first to address this discrepancy between offer and demand.

Grid users may experience inconsistent performance due to specific characteristics of grids, such as fluctuating workloads, high failure rates, and high resource heterogeneity. Although extensive research has been done in grids, providing consistent performance remains largely an unsolved problem. In this study we use overdimensioning, a simple but cost-ineffective solution, to solve the performance inconsistency problem in grids. To this end, we propose several overdimensioning strategies, and we evaluate these strategies through simulations with workloads consisting of Bag-of-Tasks. We find that although overdimensioning is a simple solution, it is a viable solution to provide consistent performance in grids.

Scientists increasingly rely on the execution of workflows in grids to obtain results from complex mixtures of applications. However, the inherently dynamic nature of grid workflow scheduling, stemming from the unavailability of scheduling information and from resource contention among the (multiple) workflows and the non-workflow system load, may lead to poor or unpredictable performance. In this paper we present a comprehensive and realistic investigation of the performance of a wide range of dynamic workflow scheduling policies in multicluster grids. We first introduce a taxonomy of grid workflow scheduling policies that is based on the amount of dynamic information used in the scheduling process, and map to this taxonomy seven such policies across the full spectrum of information use. Then, we analyze the performance of these scheduling policies through simulations and experiments in a real multicluster grid. We find that there is no single grid workflow scheduling policy with good performance across all the investigated scenarios. We also find from our real system experiments that with demanding workloads, the limitations of the head-nodes of the grid clusters may lead to performance loss not expected from the simulation results. We show that task throttling, that is, limiting the per-workflow number of tasks dispatched to the system, prevents the head-nodes from becoming overloaded while largely preserving performance, at least for communication-intensive workflows. 1.

Cloud computing has emerged as a new technology that provides large amounts of computing and data storage capacity to its users with a promise of increased scalability, high availability, and reduced administration and maintenance costs. As the use of cloud computing environments increases, it becomes crucial to understand the performance of these environments. So, it is of great importance to assess the performance of computing clouds in terms of various metrics, such as the overhead of acquiring and releasing the virtual computing resources, and other virtualization and network communications overheads. To address these issues, we have designed and implemented C-Meter, which is a portable, extensible, and easy-to-use framework for generating and submitting test workloads to computing clouds. In this paper, first we state the requirements for frameworks to assess the performance of computing clouds. Then, we present the architecture of the C-Meter framework and discuss several cloud resource management alternatives. Finally, we present our early experiences with C-Meter in Amazon EC2. We show how C-Meter can be used for assessing the overhead of acquiring and releasing the virtual computing resources, for comparing different configurations, and for evaluating different scheduling algorithms.

Large-scale distributed computing systems such as grids are serving a growing number of scientists. These environments bring about not only the advantages of an economy of scale, but also the challenges of resource and workload heterogeneity. A consequence of these two forms of heterogeneity is that job runtimes and queue wait times are highly variable, which generally reduces system performance and makes grids difficult to use by the common scientist. Predicting job runtimes and queue wait times have been widely studied for parallel environments. However, there is no detailed investigation on how the proposed prediction methods perform in grids, whose resource structure and workload characteristics are very different from those in parallel systems. In this paper, we assess the performance and benefit of predicting job runtimes and queue wait times in grids based on traces gathered from various research and production grid environments. First, we evaluate the performance of simple yet widely used time series prediction methods and the effect of applying them to different types of job classes (e.g., all jobs submitted by single users or to single sites). Then, we investigate the performance of two kinds of queue wait time prediction methods for grids. Last, we investigate whether prediction-based grid-level scheduling policies can have better performance than policies that do not use predictions.