Dependability Considerations in Distributed Control Systems
K. Žagar · 2005
Metaphorically speaking, distributed control system is a messenger that carries orders from the ’commander’ (the operator) to the ’battlefield’ (devices), and reports the other way around. As such, it is of central importance for the functioning of a large experimental physics facility. The exchanged messages must be delivered reliably: once sent, the message should reach its destination. The messenger should be highly available: whenever a message is to be sent, the sender shall not wait to be able to do so. Last but not least, the transmission should be secure: the messages should not be tampered with while en-route. In short: the control system must be dependable. In this article, ongoing research in the field of dependable distributed systems performed in the context of the 6th European Framework Programme’s DeDiSys project is presented. First, dependability issues of distributed control systems are analyzed. Then, guidelines are given on how, where and why to use techniques that improve dependability. In particular, the availability aspect of control systems is discussed, and potential for improving availability by trading it against consistency is investigated. Finally, dependability efforts we have invested in our products (in particular the ACS control system infrastructure) are described.