Thursday and Friday, December 16/17, 2004

TUD:OS '04 is co-located with the GI Betriebssysteme-Fachgruppentreffen Herbst 2004, which begins one day before TUD:OS '04, that is, on Thurday, December 16.

Please find the full program of the GI meeting on the Operating System Working Group's website.

The GI meeting's talks that are held on Friday will be presented in English. TUD:OS attendees are welcome to attend these talks as well:

Friday, 8:30 a.m.

Trusted Computing and Operating System Architectures from the Perspective of Industrial Research by Dirk Kuhlmann, HP Laboratories Bristol

Friday, 9:00 a.m.

Predictably Flexible Real-time Systems — a Scheduling Perspective by Gerhard Fohler, Department of Computer Engineering, Mälardalen University, Sweden

TUD:OS '04 – Program

Saxon State and University Library (SLUB) Dresden, Lyceum (Vortragssaal)

Friday, December 17, 2004

Friday, 10:30 a.m. – Session 1

Microkernel-based operating systems

L4Linux or Support for legacy operating systems

One of the core assumptions for our work is that systems (will) run applications with different requirements on the same machine at the same time. For example, we assume that real-time and non-real-time applications will share desktops. Or, we assume that mobile phones will be used for sensible applications with higher security requirements such as bank transfers, as well as for games and other less critical applications.

To support this scenario, many years ago we developed L4Linux, a user-level server that provides Linux kernel functionality with binary compatibility. Real-time and security-sensitive applications run besides L4Linux on a small core platform. An important aim is to reuse as much functionality residing in L4Linux for real-time and security-sensitive applications as possible. Another important aspect is the encapsulation of L4Linux such that real-time and security sensitive applications can run besides L4Linux. Future direction of work will concentrate on minimizing the changes needed for the Linux kernel to run and on providing a virtual-machine monitor that supports legacy operating systems as they are, without any changes.

Friday, 12:00 noon – Lunch

We offer complimentary lunch to all registered attendees.

Friday, 2:00 p.m. – Session 2

Microkernel-based real-time computing

First, we present a mathematical model, Quality-Assuring Scheduling (QAS), for admission and scheduling flexible applications that can tolerate occasional deadline misses. For that reason, resource requirements are split into a mandatory part that must be available and some optional parts which should be available at least with a certain percentage. The model uses the variations in the execution times of periodic applications. Thus, we can move away from worst-case reservations and drastically reduce the amount of reserved resources.

We then show how this model is mapped to different resources. For CPU, the kernel provides reservation based scheduling. We introduce the provided reservation interfaces, the feedback mechanism used to handle error conditions, and the integration with inter-process communication (IPC).

Third, the talk describes the application of QAS to the scheduling of disk requests. Disk drives impose a large variation on the execution times of disk requests, therefore real-time systems can significantly benefit from the use of statistical approaches such as QAS. The implementation of QAS has to bring in line the demand to meet the statistical guarantees with the need to optimize disk utilization. We will present our approach to solve this problem.

Another focus of our real-time research is real-time networking. Our approach applies traffic shaping techniques on top of standard Switched Ethernet technology. Real-time operating system support is needed for two things: first, to ensure timeliness of the traffic shaping process within the network drivers, and second, to ensure the application-to-application delays and bandwidth guarantees within the network stacks. Using standard Fast (and Gigabit) Ethernet network equipment, we achieve sub-millisecond application-to-application latencies with nearly full network utilization.

We will provide a live real-time demonstration of the application our models to file-system, network, and CPU scheduling. The demonstration will visualize approaches to deal with quality of service guarantees even in overload situations. It will also show how to leverage a split applications into real-time and non-real-time and into trusted and untrusted.

Friday 3:30 p.m. – Coffee Break and Demo Session

During the break, attendees have the opportunity to check out the live demonstrations of our system's real-time and security properties.

Friday 4:00 p.m. – Session 3

Microkernel-based secure computing

Our security subgroup focuses on the construction of dependable and secure systems based on our microkernel technology. Our main goal is enabling the creating of robust and trusted systems. Our approach is to combine software of different trust levels into one system. We split applications into untrusted and trusted components, which are encapsulated from each other. We allow using untrusted components for security-critical applications through introducing trusted wrappers. This approach allows us to keep the size and complexity of trusted components very small.

In our talk, we explain and demonstrate our technology using our Mikro-SINA and NIZZA projects.