The language Erlang has been developed by Ericsson to implement large switching systems. Erlang is nowadays used by several companies for complex embedded systems.The language CRL is a process algebra with data. Several verification tools are available for CRL and other process algebras, including a tool to create labelled transition systems from CRL specifications. By having a translation from Erlang to CRL we can apply the verification tools for process algebras and labelled transition systems to industrial code. The translation is aware of the major design component in the switching software. This knowledge is used to ensure that the size of the labelled transition system generated by the tools is smaller than with a naive translation.

The VoDka server is a video-on-demand system for a Spanish cable company. We look at the distributed scheduler of this system. This scheduler enables that whenever a user agent is asking for a certain movie, this request is transferred through the system and a set of possible play-back qualities is returned to the agent. In case of a non-empty set, the agent selects one and the movie is streamed to the user.The storage subsystem of the server is composed by a hierarchy of different storage systems, i.e. disks, CD players or tapes. These devices all have restrictions of which the process controlling the device is aware of. A second layer of processes controls a set of devices in one machine and has restrictions, for example, the bandwidth of its connection. A third layer may be further out in the network and serve as a cache to store more popular movies.Every process in the scheduler of the system has a function determining local restrictions, given the configuration and present state of the system. We have built a tool to construct complete models of several configurations. With techniques from the area of formal methods (in particular model checking) these models are used to determine global properties of the system, such as the maximum number of a certain class of movies that can be served in parallel.

No abstract

No abstract