APTDYN 2003: Analysis and Prediction of the Dynamic Behavior of Applications, Hosts, and Networks

These are scheduled project presentations from CS 395/495 Analysis and Prediction of the Dynamic Behavior of Applications, Hosts, and Networks, Spring 2003.

All presentations are in the large CS Classroom at 1890 Maple, on Tuesday, June 10, 2003, from 9am to noon.

All members of the Northwestern CS community (faculty, graduate students, undergraduates) are invited to attend. Presentation format is ACM conference style: 20 minutes followed by 5 minutes of questions.

  • 9:00: Rachel Goldborough, Exploring Filesystem Read/Write Prediction
    Abstract TBA.

  • 9:25: Ananth Sundararaj, Efficient Monitoring of Virtual Machines Resources

    We address the problem of efficient monitoring of virtual machines resources hosted on a physical host machine. Typically on a physical machine one uses the monitoring information available in the operating system (e.g., /proc on Linux or the performance counters in Windows) to study its performance. So given the monitoring information in the host operating system, we attempt to reconstruct the monitoring information in the guest operating systems residing on the virtual machines. The aggregate system performance is characterized using time series analysis and a mapping from the aggregate system resources to individual virtual machine system resources is developed. Such a model can be the basis to build monitoring tools for systems where many virtual machines are supported on a single physical host.

  • 9:50: Bob Adolf, Genetic Programming for Intrusion Detection

    Abstract TBA

  • 10:15: Stefan Birrer, Network Measurement System

    First I will present the prototype of a scalable network measurement system. It supports several queries for network resources. Among these are host information, host load and link rtt. I will present the architecture and the implementation of this measurement system.

    I used this system to investigate the correlation between hop distance and rtt in a planetary environment (PlanetLab). The results of this study will be presented in this talk.

  • 10:40: Anshuman Dabriwala, Genetic Programming for Time Series Prediction

    Abstract TBA

  • 11:05: Ashish Gupta and Bin Lin, How does System Resource Usage Affect User Irritation?

    In our day-to-day interactive use of desktop PCs, significant amount of system resources (like CPU, memory, network bandwidth) remain unused. We want to answer the question: How much of these resources can we utilize for other purposes, without affecting user interactivity? The user should not be irritated.

    We develop an application, which artificially stresses the system according to different patterns, and allows user feedback to express his irritation. Based on collected data, we want to develop a reliable model for this relationship. This can significantly help workstation sharing distributed systems and background applications to utilize more resources without irritating the user. It can also motivate design of interactivity-aware scheduling algorithms.