Borg Research

In a broader view we are researching how we can bring distribution in all its aspects to a higher more conceptual level, such that designing applications in wide area networks becomes easier. This includes things like as shielding the programmer from finding its communication partners, how communication takes place, how programs can migrate in a location transparent way as well as more conceptual things such as how programs in wide area networks can be made secure, how transaction and replication are managed and how distributed programs should be conceived. The main questions we ask ouselves is how all these aspects can be integrated in a programming language and how we, with such a programming language in mind, design distributed applications.

The approach we use is a bottom up approach in which we first conceived a suitable language in which we don't have to take care of aspects as serializing and routering of messages. This task is more or less finished and now we are looking in what way we can design highly structured distributed applications in a way which separates the functional requirements of the program. For example, how we can abstrahate the control flow of the program in a set of reuasable components in a practical way.

Another track of research focuses on the impact security has on distributed systems. Writing distributed systems is one, writing secure distributed systems is difficult, writing secure distributed programs in wide area networks, which are protected against there interpreting environment is probably one of the most difficult things which one can do in the field of security. With this work we hope to have a good understanding of combining distributed aspects with other aspects such as security (and eventually others, like transaction and resource management). In the end we hope to create a system and a design methodolgy which is able to take different views upon all these aspects and which is able to combine them as needed, without too much hassle.

To be able to do this kind of research we needed an architecture which gives us the abbility to write mobile multi-agents in. One can think of creating a library which allows these kind of things but the problems we faced were of the kind which didn't allowed us to write one for mobile multi-agents. We tried to do this in Java, with java RMI. After all, Java is a beautiful language (at first sight) and especially Java RMI looked very promissing because we didn't have to take care of serialising objects. The main drawback of Java (and other languages) is the unability to migrate a process to another processor. (technically speaking, we were unable to migrate a Java thread to another java virtual machine). So finally we decided to adapt an existing virtual machine to our needs.

The virtual machine we had to our disposal was Pico [http://pico.vub.ac.be/], an educational language developed by Theo D'Hondt. We choose it because it's small, we had an implementation available and students know how to use it, so we can use it to teach distributed systems with it. We named the result Borg.

If you want to know more (please fund my account 001-2797205-92 with the message 'carry on the good work' :-)

0. General Use of Mobile Agent Systems

1. The Borg Mobile Multi Agent System

2. Reflection and Computational State

3. Routing

4. Synchronization/Control Flow Glueing

5. Quality Of Services