Archer: A Community Distributed Computing Infrastructure for Computer Architecture Research and Education

📝 Abstract

This paper introduces Archer, a community-based computing resource for computer architecture research and education. The Archer infrastructure integrates virtualization and batch scheduling middleware to deliver high-throughput computing resources aggregated from resources distributed across wide-area networks and owned by different participating entities in a seamless manner. The paper discusses the motivations leading to the design of Archer, describes its core middleware components, and presents an analysis of the functionality and performance of a prototype wide-area deployment running a representative computer architecture simulation workload.

💡 Analysis

This paper introduces Archer, a community-based computing resource for computer architecture research and education. The Archer infrastructure integrates virtualization and batch scheduling middleware to deliver high-throughput computing resources aggregated from resources distributed across wide-area networks and owned by different participating entities in a seamless manner. The paper discusses the motivations leading to the design of Archer, describes its core middleware components, and presents an analysis of the functionality and performance of a prototype wide-area deployment running a representative computer architecture simulation workload.

📄 Content

1 Archer: A Community Distributed Computing Infrastructure for Computer Architecture Research and Education

Renato Figueiredo, P. Oscar Boykin, José A. B. Fortes, Tao Li,
Jie-Kwon. Peir, David Wolinsky (University of Florida) Lizy John (University of Texas at Austin) David Kaeli (Northeastern University) David Lilja (University of Minnesota) Sally McKee (Cornell University) Gokhan Memik (Northwestern University) Alain Roy (University of Wisconsin-Madison) Gary Tyson (Florida State University)

Abstract This paper introduces Archer, a community-based computing resource for computer architecture research and education. The Archer infrastructure integrates virtualization and batch scheduling middleware to deliver high-throughput computing resources aggregated from resources distributed across wide-area networks and owned by different participating entities in a seamless manner. The paper discusses the motivations leading to the design of Archer, describes its core middleware components, and presents an analysis of the functionality and performance of a prototype wide-area deployment running a representative computer architecture simulation workload.

1. Introduction Modern computer architecture research is driven by quantitative analysis. Leading-edge research requires detailed, cycle-accurate evaluation of many benchmark applications with several simulated configurations and is thus tightly dependent on the availability of high- throughput computing (HTC) systems. Many research groups are hindered in their ability to perform research because of lack of access to such resources. This is because, in addition to hardware costs, the investment of time and funds to train and educate students and staff to deploy, maintain and effectively use such systems presents a significant barrier of entry, especially for small- to medium-sized research groups. This paper describes Archer1, a community-based computing resource for computer architecture research and education. Archer integrates technologies for resource virtualization, batch job schedulers, and multi- institution collaboration, in order to create: • A computing infrastructure which scales in capacity with community buy-in: Archer starts from a seed set of cluster resources deployed at the Florida Statue University, Northeaster University, University of Texas at Austin, Northwestern University, University of Minnesota, Cornell University, and University of Florida. Subsequently, each new user joining Archer with one or more desktops or servers seamlessly contribute to its aggregate capacity. • A system that is easy for non-experts to join and use: Archer relies on packaging and distribution of software environments for HTC as self-configuring virtual networks of virtual appliances, which can easily be installed by individual users in their own resources. Surveys from users of the virtual appliance used as a basis for Archer shows that users with no prior experience can typically install and use the system within 30 minutes.

1 The Archer community infrastructure and Wiki are accessible at: http://archer-project.org

2 • A community-based repository of simulation environments: Archer allows sharing not only of hardware resources, but also of full-fledged software simulation modules consisting of application executables, support scripts, input and output data sets, and usage documents. In doing so, Archer facilitates the dissemination of useful tools and data sets, and foster creation of reproducible simulation experiments. The community-driven features in Archer provide a new way to swiftly create grids of medium size, differentiating it from related infrastructures such as the Open Science Grid (OSG) and TeraGrid, in three important ways. First, Archer enables seamless addition of resources by the community, at a fine grain (at a minimum a single desktop computer by an individual user), within minutes. This is in contrast to OSG and TeraGrid, where individual resources cannot be easily incorporated, and to gain access to resources often takes days or weeks. Second, Archer deployments are virtualized and can be easily replicated, both at a smaller scale within an institution, and at a multi-institution scale by research communities. Archer’s replicability enables research groups to easily bring up local Archer pools and be assured of preemptive access to their resources when needed, while providing opportunistic cycles to the community. This is in contrast to OSG and TeraGrid, which are large-scale shared physical resources not easily replicable at a small scale on local resources. Third, Archer empowers entry-level users to quickly learn HTC skills, from basic to advanced, with a combination of examples tailored to computer architecture and an interactiveinterface hosted on their own workstations. This is in contrast to OSG and TeraGrid, where entr

This content is AI-processed based on ArXiv data.