Student Posters

Performance Modeling based Scheduling of EMAN Workflows — by admin — last modified 2007-12-14 14:12

We present our strategies of performance-model based, plan-ahead scheduling of workflows for a bio-imaging application called EMAN [Electron Micrograph Analysis ]. We first discuss our tools and strategies for constructing computational and memory-hierarchy performance models for EMAN workflow components. We then describe our algorithm for scheduling the workflow components using the performance models. Results of our experiments show that our workflow scheduling strategies produce 1.5 to 2 times better makespan than existing strategies for this application. Also, we obtain good load balance across different Grid sites using these strategies. Contributors: Anirban Mandal, Bo Liu, Ken Kennedy, Charles Koelbel, Lennart Johnsson, Gabriel Marin, John Mellor-Crummey Presenters: Anirban Mandal, Bo Liu

A Framework for Reasoning about the Temporal Behaviors of Scientific Applications — by admin — last modified 2007-12-14 14:12

Abstract TeReSa is a general framework for expressing the qualitative temporal behavior of long-running, scientific applications for validating time varying performance. TeReSa will support virtual grids in the monitoring and validation of performance contracts for scientific workloads and in making temporal performance decisions. Poster Contributors: Emma Buneci, Daniel Reed. Poster Presented by Emma Buneci

Scalable Cross-Architecture Predictions of Memory Latency for Scientific Applications — by admin — last modified 2007-12-14 14:12

Abstract The gap between processor and memory speeds has been growing with each new generation of microprocessors. As a result, memory hierarchy response has become a critical factor limiting application performance. For this reason, we have been working to model the impact of memory access latency on program performance. This poster presents a method for characterizing an application's data access patterns in a machine independent fashion. We collect data reuse distance histograms for each memory reference in a program during repeated executions with small data inputs. Then, we model the structure and scaling of each reference's reuse distance as a function of problem size. This approach enables us to predict the number of compulsory and capacity cache misses at the instruction level for architectures and problem sizes that we did not measure. In conjunction with our reuse distance models, we use a probabilistic model to estimate the number of conflict misses for set-associative caches. We validate our approach by comparing our predictions against measurements using hardware performance counters, for several benchmarks over a large range of problem sizes. Poster Contributors: Gabriel Marin and John Mellor-Crummey Poster Presented by Gabriel Marin

GridSolve implemented on VGrADS — by admin — last modified 2007-12-14 14:12

Abstract GridSolve is a project that investigates the usage of distributed computational resources connected by computer networks to solve complex scientific problems efficiently. It is a remote procedure call (RPC)-based client/agent/server system that allows users to discover, access, and utilize remotely housed software modules, as well as the computational hardware needed to run these modules. The overall goal of integrating GridSolve with VGrADS is to provide end user multiple easy-to-use interfaces to the virtual grids. GridSolve has implemented several interfaces such as Fortran, C, Matlab, Mathematica, and Octave. By seamlessly integrating GridSolve agent with VGrADS execution system, user is allowed to access and utilize Grid resource with interface of his choice. The integrated system can also be used to extend the capabilities of problem solving environments (PSE), such as Matlab, by increasing the number and types of implemented algorithms available and solving them on Grid resources. Poster Contributors: Zhiao Shi, Asim YarKhan, Jack Dongarra Poster Presented by Zhiao Shi

Optimal Checkpoint Scheduling using Automatic Resource Characterization — by admin — last modified 2007-12-14 14:12

Abstract The architectures supported by the VGrADS initiative include high performance clusters, large scale multi-processor machines, workstation/server pools and even individual machines. For the abstraction and virtualization purposes of the VGrADS project, the components of these architectures must be statistically characterized. Although many component characteristics have been studied, one that has remained difficult to understand is that of machine availability. To serve the needs of the VGrADS infrastructure, we have developed a set of tools which automatically models and make predictions on machine availability data. Currently, we have tested our modeling and prediction techniques on three distinct resource types including student lab workstations, Internet hosts, and Condor machines. The ability to accurately model and predict resource availability durations is useful to the VGrADS software to both implement and optimize robustness mechanisms. To demonstrate, we have developed a generalized checkpoint scheduler that automatically determines a checkpoint schedule based on dynamic easurement taken from an individual resource. As Condor represents the most difficult resource pool, we have empirically tested our checkpoint scheduler on a real world Condor resource pool. Our scheduler achieves slightly better performance than previous techniques with regards to time efficiency, but has drastically improved network performance characteristics. In addition to results from our checkpointing experiments, we were able to verify a software simulation of our checkpointing system that can be used in the future to perform tests on different data sets and model parameters. In future work, we plan to enhance the entire system by employing non-parametric modeling techniques which require less data and are often more accurate. Poster Contributors Daniel Nurmi John Brevik Rich Wolski Poster Presented by Daniel Nurmi

Scheduling Compute Intensive Applications in Volatile, Shared Resource (Grid) Environments — by admin — last modified 2007-12-14 14:12

Abstract While Grid computing has become popular in recent years, understanding Grid application performance remains a challenge. In open, shared resource Grid environments, applications face heterogeneity in resources and dynamic load on the resources. Variance in runtime prediction models leads to further variability in application performance. We investigate factors affecting application performance by extensive simulation using actual compute intensive applications runtimes and actual dynamic resource load. Poster Contributors Richard Huang, Henri Casanova, Andrew Chien Poster Presented by Richard Huang

Virtual Grids: Adaptive Resource Environments for High Performance Grid Applications — by admin — last modified 2007-12-14 14:12

Abstract Grid applications are highly dependent on the description, selection and binding of the resources. The supporting middleware must be able to hide the complexity of the underlying resources, yet provide the user with the ability to control them in different granularity levels. A Virtual Grid is an abstraction that can handle complex and dynamic resource environments simplifying the development of an application. This is accomplished through our execution system (vgES) which includes a very strong and novel description language (vgDL) and an efficient resource selection and binding mechanism (vgFAB). Adaptation is also facilitated by a distributed monitoring component (vgMON), ensuring in either a transparent or application-directed fashion, that application requirements are met. Both static and dynamic resource information in vgES are provided by an intelligent agent either through batch or on-demand request. Experiments on the prototype implementation of vgFAB show that the mechanism employed is efficient and scalable with respect to request complexity, grid size and workload, providing quantifiably good matches quickly. Robustness is also verified by simulating failure and resource contention scenarios. Poster Contributors Yang-Suk Kee, Jerry Chou, Dionysios Logothetis, Richard Huang, Kenneth Yocum, Henri Casanova, Andrew A. Chien Poster Presented by Jerry Chou

Optimizing Grid-Based Workflow Execution — by admin — last modified 2007-12-14 14:12

Abstract Computational Grids provide the resources for executing large scale application that are not possible to execute on a single resources. Workflows are being used as the programming paradigm for the applications that need Grid resources. A workflow is represented as a DAG where the vertices are the compute tasks and the edges are the data dependencies between the tasks. Previous research has focussed on developing algorithms for scheduling them on Grid resources and easy to use interfaces for composing workflows and submitting them on the Grid. In this research, we focus on the behavior of the workflow execution engine using a dynamic Condor pool as an instantiation of a Distributed Virtual Computer. We show that the efficiency of the workflow execution engine in executing the workflow can make a large impact on the workflow completion times for fine granularity workflows. We examine the factors that affect the performance of the workflow execution engine and their impact on the completion time. Finally we explore some restructuring techniques that can further improve the workflow completion time. Poster Contributors Gurmeet Singh, Carl Kesselman and Ewa Deelman Poster Presented by Gurmeet Singh