The FutureGrid Experiment Management Framework * Extended Abstract Gregor von Laszewski, †‡ Geoffrey C. Fox, Fugang Wang, Andrew Younge, Archit Kulshrestha, Greg Pike Pervasive Technology Institute Indiana University Bloomington, IN 47408 laszewski@gmail.com Warren Smith, TACC Jens Vöckler, USC Renato J. Figueiredo University of Florida ABSTRACT FutureGrid provides novel computing capabilities centered around reproducible experiments. This paper describes the FutureGrid Experiment Management Framework to create and execute experiments on the FutureGrid platform. These experiments consists of the activities performed by the var- ious users of FutureGrid ranging from administrators, soft- ware developers and end users. The Experiment manage- ment system consists of software tools that record user and system actions to generate a reproducible set of tasks and resource configurations. Additionally the experiment man- agement framework can be used to share not only the ex- periment setup, but also performance information for the specific instantiation of the experiment. This makes it pos- sible to compare a variety of experiment setups and analyse the impact Grid and cloud software stacks have. Keywords Cloud, Grid, Experiments, FutureGrid 1. INTRODUCTION FutureGrid(FG) provides novel computing capabilities that will enable researchers to tackle complex research challenges related to the use and security of grids and clouds. These include topics ranging from authentication, authorization, scheduling, virtualization, middleware design, interface de- sign and cybersecurity, to the optimization of grid-enabled * This document was developed with support from the Na- tional Science Foundation (NSF) under Grant No. 0910812. † Corresponding author. ‡ Authors in no particular order. and cloud-enabled computational schemes for researchers in astronomy, chemistry, biology, engineering, atmospheric sci- ence and epidemiology. FG provides a significant and new experimental computing grid and cloud test-bed to the research community, together with user support for third-party researchers conducting ex- periments on FutureGrid. The test-bed will make it possible for researchers to con- duct experiments by submitting an experiment plan, which is then executed via a sophisticated workflow engine, pre- serving the provenance and state information necessary to reproduce the experiment. The test-bed includes a geo- graphically distributed set of heterogeneous computing sys- tems, a data management system that will hold both meta- data and a growing library of software images necessary for Cloud computing, and a dedicated network allowing iso- lated, secure experiments. The test-bed will support virtual machine-based environments, as well as operating systems on native hardware for experiments aimed at minimizing overhead and maximizing performance. The project part- ners will integrate existing open-source software packages to create an easy-to-use software environment that supports the instantiation, execution and recording of grid and cloud computing experiments. One of the goals of the project is to understand the behavior and utility of cloud computing ap- proaches. Recently, cloud computing has become quite pop- ular and a multitude of cloud computing middleware have been developed. However, it is not clear at this time which of these toolkits will become the users’ choice toolkit. FG provides the ability to compare these frameworks with each other while considering real scientific applications. Hence, researchers will be able to measure the overhead of cloud technology by requesting linked experiments on both vir- tual and bare-metal systems, providing them valuable infor- mation that will help them decide which infrastructure suits them better and also help users that want to transition from one environment to the other. Participants. The list of participants to build the Future- Grid capacity includes (see Figure ??) Indiana University, Purdue University, San Diego Supercomputer Center at Uni-