Katarzyna Keahey

Virtual workspaces: Achieving quality of service and quality of life in the Grid

By defining standardized protocols for discovering, accessing, monitoring, and managing remote computers, storage systems, networks, and other resources, Grid technologies make it possible—in principle—to allocate resources to applications dynamically, in an on-demand fashion [1]. However, while Grids offer users access to many diverse and powerful resources, they do little to ensure that once a resource is accessed, it fulfills user expectations for

Sky Computing

Infrastructure-as-a-service (IaaS) cloud computing is revolutionizing how we approach computing. Compute resource consumers can eliminate the expense inherent in acquiring, managing, and operating IT infrastructure and instead lease resources on a pay-as-you-go basis. IT infrastructure providers can exploit economies of scale to mitigate the cost of buying and operating resources and avoid the complexity required to manage multiple customer-specific environments and applications. The authors describe the context in which cloud computing arose, discuss its current strengths and shortcomings, and point to an emerging computing pattern it enables that they call sky computing.

Virtual workspaces in the grid

Despite significant progress in the development of Grid infrastructure, the provisioning of a customized and controllable remote execution environment remains an open issue. This paper introduces the concept of a virtual workspace, a configurable execution environment that can be created and managed as a first-class entity to reflect client requirements. Such workspaces can be dynamically deployed on a variety of resources decoupling the notion of environment and resource. We show how virtual workspaces fit into the Grid architecture, present an example implementation using virtual machines, and discuss our initial experiences using this system in practice and with applications.

From sandbox to playground: dynamic virtual environments in the grid

Much experience has been gained with the protocols and mechanisms needed for discovery and allocation of remote computational resources. However, the preparation of a remote computer for use by a distributed application also requires the creation of an appropriate execution environment, which remains an ad hoc and often clumsy process. We propose here a codification of the interactions required to negotiate the creation of new execution environments. In brief we model dynamic virtual environments (DVEs) as first-class entities in a distributed environment, with grid service interfaces defined to negotiate creation, monitor properties, and manage lifetime. We also show how such DVEs can be implemented in a variety of technologies - sandboxes, virtual machines, or simply Unix accounts - and evaluate costs associated with these different approaches. DVEs provide a basis for both customization of a remote computer to meet user needs and also enforcement of resource usage and security policies. They can also simplify the administration of virtual organizations (VOs), by allowing new environments to be created automatically, subject to local and VO policy. Thus, DVEs have the potential to relieve much of the current administrative burden involved in providing and using grid resources.

Fine-Grain Authorization for Resource Management in the Grid Environment

In this document we describe our work-in-progress for enabling finegrain authorization of resource management. In particular we address the needs of Virtual Organizations (VOs) to enforce their own polices in addition to those of the resource owners.

PAWS: collective interactions and data transfers

The authors discuss problems and solutions pertaining to the interaction of components representing parallel applications. We introduce the notion of a collective port which is an extension of the Common Component Architecture (CCA) ports and allows collective components representing parallel applications to interact as one entity. We further describe a class of translation components, which translate between the distributed data format used by one parallel implementation to that used by another. A well known example of such components is the MxN component which translates between data distributed on M processors to data distributed on N processors. We describe its implementation in Parallel Application Work Space (PAWS), as well as the data structures PAWS uses to support it. We also present a mechanism allowing the framework to invoke this component on the programmers behalf whenever such translation is necessary, freeing the programmer from treating collective component interactions as a special case. In doing that, we introduce framework-based, user-defined distributed type casts. Finally, we discuss our initial experiments in building optimized complex translation components out of atomic functionalities.

Flying Low: Simple Leases with Workspace Pilot

As the use of virtual machines (VMs) for scientific applications becomes more common, we encounter the need to integrate VM provisioning models into the existing resource management infrastructure as seamlessly as possible. To address such requirements, we describe an approach to VM management that uses multi-level scheduling to integrate VM provisioning into batch schedulers such as PBS. We then evaluate our approach on the TeraPort cluster at the University of Chicago.

Providing data transfer with QoS as agreement-based service

Over the last decade, grids have become a successful tool for providing distributed environments for secure and coordinated execution of applications. The successful deployment of many realistic applications in such environments on a large scale has motivated their use in experimental science [L. C. Pearlman et al., (2004), K. Keahey et al. (2004)] where grid-based computations are used to assist in ongoing experiments. In such scenarios, quality of service (QoS) guarantees on execution as well as data transfer is desirable. The recently proposed WS-Agreement model [K. Czajkowski et al. K. Keahey et al. (2004)] provides an infrastructure within which such quality of service can be negotiated and obtained. We have designed and implemented a data transfer service that exposes an interface based on this model and defines agreements which guarantee that, within a certain confidence level, file transfer can be completed under a specified time. The data transfer service accepts a clients request for data transfer and makes an agreement with the client based on QoS metrics (such as the transfer time and confidence level with which the service can be provided). In our approach we use prediction as a base for formulating an agreement with the client, and we combine prediction and rate limiting to adoptively ensure that the agreement is met.

Trading Grid services within the UK e-science Grid

The Open Grid Services Architecture (OGSA) presents the Grid community with an opportunity to define standard service interfaces to enable the construction of an interoperable Grid infrastructure. The provision of this infrastructure has, to date, come from the donation of time and effort from the research community primarily for their Own use. The growing involvement of industry and commerce in Grid activity is accelerating the need to find business models that support their participation. It is therefore essential that an economic infrastructure be incorporated into the OGSA to support economic transactions between service providers and their clients. This chapter describes current standardization efforts taking place with the Global Grid Forum and the implementation of such an architecture within the UK e-Science Programme through the Computational Markets project.

The taming of the Grid : virtual application services.

In this report we develop a view of the Grid based on the application service provider (ASP) model. This view enables the user to see the Grid as a collection of application services that can be published, discovered, and accessed in a relatively straightforward manner, hiding much of the complexity involved in using computational Grids and thus making it simpler and more accessible to a wider range of users. However, in order to satisfy the requirements of real-time scientific application clients, we combine the ASP model with representation of quality of service about the execution of services and the results they produce. Specifically, we focus on real-time, deadline-bound execution as the quality of service derived by a client. We describe an architecture implementing these ideas and the role of client and server in the context of the functionality we develop. We also describe preliminary experiments using an equilibrium fitting application for magnetic fusion in our architecture.