Ronald J. Desmarais

A Design Space for Self-Adaptive Systems

Self-adaptive systems research is expanding as systems professionals recognize the importance of automation for managing the growing complexity, scale, and scope of software systems. The current approach to designing such systems is ad hoc, varied, and fractured, often resulting in systems with parts of multiple, sometimes poorly compatible designs. In addition to the challenges inherent to all software, this makes evaluating, understanding, comparing, maintaining, and even using such systems more difficult. This paper discusses the importance of systematic design and identifies the dimensions of the self-adaptive system design space. It identifies key design decisions, questions, and possible answers relevant to the design space, and organizes these into five clusters: observation, representation, control, identification, and enacting adaptation. This characterization can serve as a standard lexicon, that, in turn, can aid in describing and evaluating the behavior of existing and new self-adaptive systems. The paper also outlines the future challenges for improving the design of self-adaptive systems.

Monitoring in adaptive systems using reflection

Continuous evolution is a key trait of software-intensive systems. Many research projects investigate mechanisms to adapt software systems effectively in order to ease evolution. By observing its internal state and surrounding context continuously using feedback loops, an adaptive system is able to analyze its effectiveness by evaluating quality criteria and then self-tune to improve its operations. The goals of these feedback loops range from keeping single variables in a prescribed range to satisfying non-functional requirements by regulating decentralized, interdependent subsystems. To be able to observe and possibly orchestrate continuous evolution of software systems in a complex and changing environment, we need to push monitoring of evolving systems to unprecedented levels. It has been established that security has to be built into a system from the ground up and cannot be added as an afterthought - the same is probably true for intensive monitoring. We propose to monitor adaptive systems with autonomic elements to enhance their assessment capabilities. In this paper, we discuss how to build monitoring into Java programs from the ground up with reflection technology to detect normal and exceptional system behavior.

Deploying HEP applications using Xen and Globus Virtual Workspaces

The deployment of HEP applications in heterogeneous grid environments can be challenging because many of the applications are dependent on specific OS versions and have a large number of complex software dependencies. Virtual machine monitors such as Xen could be used to package HEP applications, complete with their execution environments, to run on resources that do not meet their operating system requirements. Our previous work has shown HEP applications running within Xen suffer little or no performance penalty as a result of virtualization. However, a practical strategy is required for remotely deploying, booting, and controlling virtual machines on a remote cluster. One tool that promises to overcome the deployment hurdles using standard grid technology is the Globus Virtual Workspaces project. We describe strategies for the deployment of Xen virtual machines using Globus Virtual Workspace middleware that simplify the deployment of HEP applications.

Status of the Raven MOAO science demonstrator

Raven is a Multi-Object Adaptive Optics (MOAO) scientific demonstrator which will be used on-sky at the Subaru observatory. Raven is currently being built at the University of Victoria AO Lab. In this paper, we present an overview of the final Raven design and then describe lab tests involving prototypes of Raven subsystems. The final design includes three open loop wavefront sensors (WFSs), a laser guide star WFS and two figure/truth WFSs. Two science channels, each containing a deformable mirror (DM), feed light to the Subaru IRCS spectrograph. Central to the Raven MOAO system is a Calibration Unit (CU) which contains multiple sources, a telescope simulator including two rotating phase screens and a ground layer DM that can be used to calibrate and test Raven. We are working with the Raven CU and open loop WFSs to test and validate our open loop calibration and alignment techniques.

Research computing in a distributed cloud environment

The recent increase in availability of Infrastructure-as-a-Service (IaaS) computing clouds provides a new way for researchers to run complex scientific applications. However, using cloud resources for a large number of research jobs requires significant effort and expertise. Furthermore, running jobs on many different clouds presents even more difficulty. In order to make it easy for researchers to deploy scientific applications across many cloud resources, we have developed a virtual machine resource manager (Cloud Scheduler) for distributed compute clouds. In response to a users job submission to a batch system, the Cloud Scheduler manages the distribution and deployment of user-customized virtual machines across multiple clouds. We describe the motivation for and implementation of a distributed cloud using the Cloud Scheduler that is spread across both commercial and dedicated private sites, and present some early results of scientific data analysis using the system.

Towards software-adaptive green computing based on server power consumption

With the proliferation of virtualization and cloud comput- ing, optimizing the power usage effectiveness of enterprise data centers has become a laudable goal and a critical re- quirement in IT operations all over the world. While a sig- nificant body of research exists to measure, monitor, and control the greenness level of hardware components, signif- icant research efforts are needed to relate hardware energy consumption to energy consumption due to program exe- cution. In this paper we report on our investigations to characterize power consumption profiles for different types of compute and memory intensive software applications. In particular, we focus on studying the effects of CPU loads on the power consumption of compute servers by monitoring rack power consumption in a data center. We conducted a series of experiments with a variety of processes of differ- ent complexity to understand and characterize the effect on power consumption. Combining processes of varying com- plexity with varying resource allocations produces different energy consumption levels. The challenge is to optimize pro- cess orchestration based on a power consumption framework to accrue energy savings. Our ultimate goal is to develop smart adaptive green computing techniques, such as adap- tive job scheduling and resource provisioning, to reduce over- all power consumption in data centers or clouds.

Characterizing problems for realizing policies in self-adaptive and self-managing systems

Self-adaptive and self-managing systems optimize their own behaviour according to high-level objectives and constraints. One way for administrators to specify goals for such optimization problems effectively is using policies. Over the past decade, researchers produced various approaches, models and techniques for policy specification in different areas including distributed systems, communications networks, web services, autonomic computing, and cloud computing. Research challenges range from characterizing policies for ease of specification in particular application domains to categorizing policies for achieving solution qualities for particular algorithmic techniques. The contributions of this paper are threefold. Firstly, we give a mathematical formulation for each of the three policy types, action, goal and utility function policies, introduced in the policy framework by Kephart and Walsh. In particular, we introduce a first precise characterization of goal policies for optimization problems. Secondly, this paper introduces a mathematical framework that adds structure to the underlying optimization problem for different types of policies. Structure is added either to the objective function or the constraints of the optimization problem. These mathematical structures, imposed on the underlying problem, progressively increase the quality of the solutions obtained when using the greedy optimization technique. Thirdly, we show the applicability of our framework by analyzing several optimization problems encountered in self-adaptive and selfmanaging systems, such as resource allocation, quality of service management, and SLA profit optimization to provide quality guarantees for their solutions. Our approach is based on the algorithmic frameworks by Edmonds, Fisher et al., and Mestre, and the policy framework of Kephart and Walsh. Our characterization and approach will help designers of self-adaptive and self-managing systems formulate optimization problems, decide on algorithmic strategies based on policy requirements, and reason about solution qualities.

BaBar MC production on the Canadian grid using a web services approach

The present paper highlights the approach used to design and implement a web services based BaBar Monte Carlo (MC) production grid using Globus Toolkit version 4. The grid integrates the resources of two clusters at the University of Victoria, using the ClassAd mechanism provided by the Condor-G metascheduler. Each cluster uses the Portable Batch System (PBS) as its local resource management system (LRMS). Resource brokering is provided by the Condor matchmaking process, whereby the job and resource attributes are expressed as ClassAds. The important features of the grid are automatic registering of resource ClassAds to the central registry, ClassAds extraction from the registry to the metascheduler for matchmaking, and the incorporation of input/output file staging. Web-based monitoring is employed to track the status of grid resources and the jobs for an efficient operation of the grid. The performance of this new grid for BaBar jobs, and the existing Canadian computational grid (GridX1) based on Globus Toolkit version 2 is found to be consistent.

The GridX1 computational Grid: from a set of service-specific protocols to a service-oriented approach

GridXl is a computational grid designed and built to link resources at a number of research institutions across Canada. Building upon the experience of designing, deploying and operating the first generation of GridXl, we have designed a second-generation, Web-services-based, computational grid. The second generation of GridXl leverages the Web services resource framework, implemented by the Globus Toolkit version 4. The value added by GridXl includes metascheduling, file staging, resource registry and resource monitoring.

A Proposal for an Autonomic Grid Management System

This paper presents a proposal for an autonomic grid management system (AGMS) using IBMs Autonomic Toolkit. The AGMS proposed would be knowledgeable on grid workflow patterns be able to sense and change grid resources and services according to a set of grid policies. The AGMS will perform preventative and corrective maintenance procedures on the grid to ensure its workflows are being completed properly. This should result in improved grid service reliability and job submission resiliency.