Dejan S. Milojicic

Adaptive offloading for pervasive computing

Delivering a complex application on a resource-constrained mobile device is challenging. An adaptive offloading system enables dynamic partitioning of the application and efficient offloading of part of its execution to a nearby surrogate. To deliver pervasive services without modifying the application or degrading its fidelity, we propose an adaptive offloading system that includes two key parts: a distributed offloading platform and an offloading inference engine. There are two important decision-making problems for adaptive offloading: adaptive offloading triggering and efficient application partitioning.

OpenNebula: A Cloud Management Tool

In this installment of Trend Wars, we discuss cloud computing and OpenNebula with Ignacio M. Llorente and Rubén S. Montero, who are the principal investigator and the chief architect, respectively, of the open source OpenNebula project.

Trend Wars - Mobile agent applications

In what application domains do mobile agents have potential deployment? I basically see three different domains. One is data-intensive applications where the data is remotely located, is owned by the remote service provider, and the user has specialized needs. Here, the user sends an agent to the server storing the data. The second domain is where agents are launched by an appliance—for example, shipping an agent from a cellular phone to a remote server. The third and maybe most important is for extensible servers, where a user can ship and install an agent representing him more permanently on a remote server. The agent is now a personalized, autonomous piece of code that runs remotely and only contacts the user whenever events of interest to the user occur.

Open Cirrus: A Global Cloud Computing Testbed

Open Cirrus is a cloud computing testbed that, unlike existing alternatives, federates distributed data centers. It aims to spur innovation in systems and applications research and catalyze development of an open source service stack for the cloud.

Towards a distributed platform for resource-constrained devices

Many visions of the future predict a world with pervasive computing, where computing services and resources permeate the environment. In these visions, people will want to execute a service on any available device without worrying about whether the service has been tailored for the device. We believe that it will be difficult to create services that can execute well on the wide variety of devices that are being developed because of problems with diversity and resource constraints. We believe that these problems can be greatly reduced by using an ad-hoc distributed platform to transparently off-load portions of a service from a resource-constrained device to a nearby server. We implemented a preliminary prototype and emulator to study this approach. Our experiments show the beneficial use of nearby resources to relieve both memory and processing constraints, when it is appropriate to do so. We believe that this approach will reduce the burden on developers by masking more device details.

Adaptive offloading inference for delivering applications in pervasive computing environments

Pervasive computing allows a user to access an application on heterogeneous devices continuously and consistently. However it is challenging to deliver complex applications on resource-constrained mobile devices, such as cellular telephones and PDA. Different approaches, such as application-based or system-based adaptations, have been proposed to address the problem. However existing solutions often require degrading application fidelity. We believe that this problem can be overcome by dynamically partitioning the application and offloading part of the application execution to a powerful nearby surrogate. This will enable pervasive application delivery to be realized without significant fidelity degradation or expensive application rewriting. Because pervasive computing environments are highly dynamic, the runtime offloading system needs to adapt to both application execution patterns and resource fluctuations. Using the fuzzy control model, we have developed an offloading inference engine to adaptively solve two key decision-making problems during runtime offloading: (1) timely triggering of adaptive offloading, and (2) intelligent selection of an application partitioning policy. Extensive trace-driven evaluations show the effectiveness of the offloading inference engine.

SLA Decomposition: Translating Service Level Objectives to System Level Thresholds

In todays complex and highly dynamic computing environments, systems/services have to be constantly adjusted to meet service level agreements (SLAs) and to improve resource utilization, thus reducing operating cost. Traditional design of such systems usually involves domain experts who implicitly translate service level objectives (SLOs) specified in SLAs to system-level thresholds in an ad-hoc manner. In this paper, we present an approach that combines performance modeling with performance profiling to create models that translate SLOs to lower-level resource requirements for each system involved in providing the service. Using these models, the process of creating an efficient design of a system/service can be automated, eliminating the involvement of domain experts. We demonstrate that our approach is practical and that it can be applied to different applications and software architectures. Our experiments show that for a typical 3-tier e-commerce application in a virtualized environment the SLAs can be met while improving CPU utilization up to 3 times.

Evaluation of HPC Applications on Cloud

HPC applications are increasingly being used in academia and laboratories for scientific research and in industries for business and analytics. Cloud computing offers the benefits of virtualization, elasticity of resources and elimination of cluster setup cost and time to HPC applications users. However, poor network performance, performance variation and OS noise are some of the challenges for execution of HPC applications on Cloud. In this paper, we propose that Cloud can be viable platform for some HPC applications depending upon application characteristics such as communication volume and pattern and sensitivity to OS noise and scale. We present an evaluation of the performance and cost tradeoffs of HPC applications on a range of platforms varying from Cloud (with and without virtualization) to HPC-optimized cluster. Our results show that Cloud is viable platform for some applications, specifically, non communicationintensive applications such as embarrassingly parallel and tree-structured computations up to high processor count and for communication-intensive applications up to low processor count.

Mobile Objects and Agents (MOA)

This paper describes the design and implementation of the Mobile Objects and Agents (MOA) project at the Open Group Research Institute. MOA was designed to support migration, communication and control of agents. It was implemented on top of the Java Virtual Machine, without any modifications to it. The initial project goals were to support communication across agent migration, as a means for collaborative work; and to provide extensive resource control, as a basic support for countering denial of service attacks. In the course of the project we added two further goals: compliance with the Java Beans component model which provides for additional configurability and customization of agent system and agent applications; and interoperability which allows cooperation with other agent systems. This paper analyzes the architecture of MOA, in particular the support for mobility, naming and locating, communication, and resource management. Object and component models of MOA are discussed and some implementation details described. We summarize the lessons learned while developing and implementing MOA and compare it to related work.

Susceptibility of commodity systems and software to memory soft errors

It is widely understood that most system downtime is accounted for by programming errors and administration time. However, a growing body of work has indicated an increasing cause of downtime may stem from transient errors in computer system hardware due to external factors, such as cosmic rays. This work indicates that moving to denser semiconductor technologies at lower voltages has the potential to increase these transient errors. In this paper, we investigate the susceptibility of commodity operating systems and applications on commodity PC processors to these soft-errors and we introduce ideas regarding the improved recovery from these transient errors in software. Our results indicate that, for the Linux kernel and a Java virtual machine running sample workloads, many errors are not activated, mostly due to overwriting. In addition, given current and upcoming microprocessor support, our results indicate that those errors activated, which would normally lead to system reboot, need not be fatal to the system if software knowledge is used for simple software recovery. Together, they indicate the benefits of simple memory soft error recovery handling in commodity processors and software.