|Final Report of the COST-247 Action|
Developments in the definition and verification of the Xpress Transfer Protocol (XTP)
|Institut National des Telecommunications (INT)|
|9, rue Charles Fourier|
|91011 EVRY cedex|
Description. XTP was originally designed as a transfer layer protocol (transport and network layers)[PEI92]. XTP functions can be dynamically configured, using a minimal and flexible set of simple mechanisms ("light-weight transport"), in order to offer efficient solutions for different user requirements and network characteristics. Significant changes of the protocol definition were decided at the last meeting of the XTP Forum (the organisation that promotes the protocol), in July 1994. The new revision of the protocol definition (4.0), restricts XTP to the transport layer and is still under development.
Aplication. XTP was aimed to provide high performance for a broad range of service requirements, in high speed networks: it offers support for distributed, transactional and multimedia applications and efficient transfer of bulk data. It is currently used in experimental distributed multimedia systems, in USA and Europe.
Validation technique. The protocol was specified in Estelle, according to the previous definition (3.6) and the brief description of the modifications, distributed after the meeting of the Forum. No service definition is available. The verification was made using a generic transport service, defined for this purpose, and a connectionless network service. The generic service provides control of the configuration options and allows to take advantage of the XTP features, for performing efficiently different service types. A configuration consisting of XTP entities and a model of its environment was extensively simulated for a typical set of scenarios, using Edb, the simulation tool of the Estelle Development Toolset [Budk92], [EDT]. The fast and systematic production of scenarios was facilitated by the design of a "programmable" user, configured by Edb macros and observers. Verification automata, implemented with Edb observers, were used for checking individual mechanisms and the offered service, during the simulations.
Conclusion. This study was aimed to offer support to the current conception phase, by exploring the effects of the modifications and by identifying and evaluating solutions for the new situations that might appear. It showed that the possibility to shape the protocol mechanisms by observing and verifying the simulations, interactively and automatically, can be of great help during protocol design. The study also offered the opportunity to improve the methods for automatic verification during the simulations, working on a real-life protocol, and to identify functions to be added to Edb, in order to facilitate the implementation of the observers.
This presentation has been given during the COST-247 3rd Management Committee Meeting (Evry, France, September 19--20, 1994).
COST-247 Working Group(s): 2