Kate Keahey

On the Use of Cloud Computing for Scientific Workflows

This paper explores the use of cloud computing for scientific workflows, focusing on a widely used astronomy application-Montage. The approach is to evaluate from the point of view of a scientific workflow the tradeoffs between running in a local environment, if such is available, and running in a virtual environment via remote, wide-area network resource access. Our results show that for Montage, a workflow with short job runtimes, the virtual environment can provide good compute time performance but it can suffer from resource scheduling delays and widearea communications.

Elastic Site: Using Clouds to Elastically Extend Site Resources

Infrastructure-as-a-Service (IaaS) cloud computing offers new possibilities to scientific communities. One of the most significant is the ability to elastically provision and relinquish new resources in response to changes in demand. In our work, we develop a model of an “elastic site” that efficiently adapts services provided within a site, such as batch schedulers, storage archives, or Web services to take advantage of elastically provisioned resources. We describe the system architecture along with the issues involved with elastic provisioning, such as security, privacy, and various logistical considerations. To avoid over- or under-provisioning the resources we propose three different policies to efficiently schedule resource deployment based on demand. We have implemented a resource manager, built on the Nimbus toolkit to dynamically and securely extend existing physical clusters into the cloud. Our elastic site manager interfaces directly with local resource managers, such as Torque. We have developed and evaluated policies for resource provisioning on a Nimbus-based cloud at the University of Chicago, another at Indiana University, and Amazon EC2. We demonstrate a dynamic and responsive elastic cluster, capable of responding effectively to a variety of job submission patterns. We also demonstrate that we can process 10 times faster by expanding our cluster up to 150 EC2 nodes.

Overhead Matters: A Model for Virtual Resource Management

Virtual machines provide a promising vehicle for controlled sharing of physical resources, allowing us to instantiate a precisely defined virtual resource, configured with desired software configuration and hardware properties, on a set of physical resources. We describe a model of virtual machine provisioning in a Grid environment that allows us to define such virtual resources and efficiently instantiate them on a physical Grid infrastructure. We argue that to properly account for, and manage, the overhead resulting from instantiating and managing virtual resources, overhead must be scheduled at the same level as virtual resources, instead of being deducted from a users resource allocation. We present preliminary results that demonstrate the benefits of such an approach.

Improving Utilization of Infrastructure Clouds

A key advantage of infrastructure-as-a-service (IaaS) clouds is providing users on-demand access to resources. To provide on-demand access, however, cloud providers must either significantly overprovision their infrastructure (and pay a high price for operating resources with low utilization) or reject a large proportion of user requests (in which case the access is no longer on-demand). At the same time, not all users require truly on-demand access to resources. Many applications and workflows are designed for recoverable systems where interruptions in service are expected. For instance, many scientists utilize high-throughput computing (HTC)-enabled resources, such as Condor, where jobs are dispatched to available resources and terminated when the resource is no longer available. We propose a cloud infrastructure that combines on-demand allocation of resources with opportunistic provisioning of cycles from idle cloud nodes to other processes by deploying backfill virtual machines (VMs). For demonstration and experimental evaluation, we extend the Nimbus cloud computing toolkit to deploy backfill VMs on idle cloud nodes for processing an HTC workload. Initial tests show an increase in IaaS cloud utilization from 37.5% to 100% during a portion of the evaluation trace but only 6.39% overhead cost for processing the HTC workload. We demonstrate that a shared infrastructure between IaaS cloud providers and an HTC job management system can be highly beneficial to both the IaaS cloud provider and HTC users by increasing the utilization of the cloud infrastructure (thereby decreasing the overall cost) and contributing cycles that would otherwise be idle to processing HTC jobs.

Design of the FutureGrid experiment management framework

FutureGrid provides novel computing capabilities that enable reproducible experiments while simultaneously supporting dynamic provisioning. This paper describes the FutureGrid experiment management framework to create and execute large scale scientific experiments for researchers around the globe. The experiments executed are performed by the various users of FutureGrid ranging from administrators to software developers and end users. The Experiment management framework will consist of software tools that record user and system actions to generate a reproducible set of tasks and resource configurations. Additionally, the experiment management framework can be used to share not only the experiment setup, but also performance information for the specific instantiation of the experiment. This makes it possible to compare a variety of experiment setups and analyze the impact Grid and Cloud software stacks have.

Virtual workspaces for scientific applications

Scientists often face the need for more computing power than is available locally, but are constrained by the fact that even if the required resources were available remotely, their complex software stack would not be easy to port to those resources. Many applications are dependency-rich and complex, making it hard to run them on anything but a dedicated platform. Worse, even if the applications do run on another system, the results they produce may not be consistent across different runs. As part of the Center for Enabling Distributed Petascale Science (CEDPS) project we have been developing the workspace service which allows authorized clients to dynamically provision execution environments, using virtual machine technology, on remote computers. Virtual machines provide an excellent implementation of a portable environment as they allow users to configure an environment and then deploy it on a variety of platforms. This paper describes a proof-of-concept of this strategy developed for the High-Energy and Nuclear Physics (HENP) applications such as STARs. We are currently building on this work to enable production STAR runs in virtual machines.

Going back and forth: efficient multideployment and multisnapshotting on clouds

Infrastructure as a Service (IaaS) cloud computing has revolutionized the way we think of acquiring resources by introducing a simple change: allowing users to lease computational resources from the cloud providers datacenter for a short time by deploying virtual machines (VMs) on these resources. This new model raises new challenges in the design and development of IaaS middleware. One of those challenges is the need to deploy a large number (hundreds or even thousands) of VM instances simultaneously. Once the VM instances are deployed, another challenge is to simultaneously take a snapshot of many images and transfer them to persistent storage to support management tasks, such as suspend-resume and migration. With datacenters growing rapidly and configurations becoming heterogeneous, it is important to enable efficient concurrent deployment and snapshotting that are at the same time hypervisor independent and ensure a maximum compatibility with different configurations. This paper addresses these challenges by proposing a virtual file system specifically optimized for virtual machine image storage. It is based on a lazy transfer scheme coupled with object versioning that handles snapshotting transparently in a hypervisor-independent fashion, ensuring high portability for different configurations. Large-scale experiments on hundreds of nodes demonstrate excellent performance results: speedup for concurrent VM deployments ranges from a factor of 2 up to 25, with a reduction in bandwidth utilization of as much as 90%.

A CORBA Commodity Grid Kit

This paper reports on an ongoing research project aimed at designing and deploying a Common Object Resource Broker Architecture (CORBA) (ww.omg.org) Commodity Grid (CoG) Kit. The overall goal of this project is to enable the development of advanced Grid applications while adhering to state‐of‐the‐art software engineering practices and reusing the existing Grid infrastructure. As part of this activity, we are investigating how CORBA can be used to support the development of Grid applications. In this paper, we outline the design of a CORBA CoG Kit that will provide a software development framework for building a CORBA ‘Grid domain’. We also present our experiences in developing a prototype CORBA CoG Kit that supports the development and deployment of CORBA applications on the Grid by providing them access to the Grid services provided by the Globus Toolkit. Copyright

Managing appliance launches in infrastructure clouds

Infrastructure cloud computing introduces a significant paradigm shift that has the potential to revolutionize how scientific computing is done. However, while it is actively adopted by a number of scientific communities, it is still lacking a well-developed and mature ecosystem that will allow the scientific community to better leverage the capabilities it offers. This paper introduces a specific addition to the infrastructure cloud ecosystem: the cloudinit.d program, a tool for launching, configuring, monitoring, and repairing a set of interdependent virtual machines in an infrastructure-as-a-service (IaaS) cloud or over a set of IaaS clouds. The cloudinit.d program was developed in the context of the Ocean Observatory Initiative (OOI) project to help it launch and maintain complex virtual platforms provisioned on demand on top of infrastructure clouds. Like the UNIX init.d program, cloudinit.d can launch specified groups of services and the VMs in which they run, at different run levels representing dependencies of the launched VMs. Once launched, cloudinit.d monitors the health of each running service to ensure that the overall application is operating properly. If a problem is detected in a service, cloudinit.d will restart only that service and any other service that failed that depended on it.

Cumulus: an open source storage cloud for science

Amazons S3 protocol has emerged as the de facto interface for storage in the commercial data cloud. However, it is closed source and unavailable to the numerous science data centers all over the country. Just as Amazons Simple Storage Service (S3) provides reliable data cloud access to commercial users, scientific data centers must provide their users with a similar level of service. Ideally scientific data centers could allow the use of the same clients and protocols that have proven effective to Amazons users. But how well does the S3 REST interface compare with the data cloud transfer services used in todays computational centers? Does it have the features needed to support the scientific community? If not, can it be extended to include these features without loss of compatibility? Can it scale and distribute resources equally when presented with common scientific the usage patterns? We address these questions by presenting Cumulus, an open source implementation of the Amazon S3 REST API. It is packaged with the Nimbus IaaS toolkit and provides scalable and reliable access to scientific data. Its performance compares favorably with that of GridFTP and SCP, and we have added features necessary to support the econometrics important to the scientific community.