Analysis of real-time properties of open distributed systems
Research on real-time systems has resulted in many effective
technologies enabling end-to-end performance requirements of
a system to be studied as a central problem. This analysis is typically
done based on resource utilization and external demands on each computing
process. This project studies how real-time techniques
can be incorporated into open system architectures
in development platforms such as CORBA. The goal is to facilitate trade-off
studies at early design stage, and support component based
software development.
The component based approach in combination with
applications with clear real-time and dependability demands poses substantial
requirements on CORBA based systems. These are of two main categories:
demands on the operating environment (operating system, underlying network)
and demands on the CORBA runtime system (guarantees or exceptions,
time handling, event handling, priority handling).
On top of this, every reused component must come with declarations of worst
case bounds and clear use patterns (design rules) to contribute to a desired
global effect. Extending existing distributed system development environments
with such capabilities is an active area for the research community.
A major application area for such technologies is in
telecommunication systems. These systems have high availability requirements,
and major components in the system architecture are devoted to operation
monitoring. As well as operator-based configuration management there are
elements of self adjustment and replication to achieve fault-tolerance.
On the Fault-tolerance side we study the basic primitives for building
replication services where groups of processes are employed to achieve
fault-tolerance.
To study the real-time and fault-tolerance trade-offs in this setting
is a major goal of our research.
During the first phase of the project we started to investigate
the combination of existing real-time scheduling
and resource management technologies with the emerging component based
development infrastructures for distributed systems. This is to support the
next stage where we want to build extensions to existing systems engineering
development tools so that the above mentioned trade-offs can be studied
in more detail. The result of the survey over approaches to extend CORBA with
real-time has resulted in a 24 page draft report (available from
http://www.ida.liu.se/~snt/research/RT-CORBA-survey.pdf). An abridged version of this
survey has been submitted to an international conference for review.
Graduate students: Calin Curescu, Diana Szentivanyi
Supervisor: Simin Nadjm-Tehrani
