Tsuyoshi Muramatsu

A data-driven implementation of telecommunication network systems

Undoubtedly, users requirements and engineering constraints will influence the future information environment. The authors have started a project named CUE (Coordinating Users requirements and Engineering constraints). The project uses an efficient ultra-multiprocessing capability with overload tolerance and maintainability/expandability, incorporating run-time reconfiguration as essential requirements to realize distributed networking systems such as TINA, which is aimed at the interoperability, reusability and transparency of application software. To satisfy these requirements from engineering viewpoints, the CUE project proposes applying to the network systems an autonomous decentralized data-driven processor which adopts an autonomous execution control scheme in which data packets passing through the pipeline achieve all data transfer, processing and storing/buffering functions in a completely decentralized manner. The processor can therefore perform high flow-rate multimedia streams with considerable flow fluctuations and act as a functional module flexibly adding to/removing from the system at run-time without any overheads. The essential requirements to realize a distributed networking environment such as TINA are first discussed. The applicability of our data-driven processor is shown through actual performance evaluations by the latest VLSI systems. An autonomous run-time reconfiguration scheme in a multiprocessor system is proposed to demonstrate the potential of the data-driven processor. Finally, the current status and future directions of the CUE project are briefly addressed.

Data-driven implementation of TINA kernel Transport Network

To realize the actual TINA-based telecommunications network, the performance of the kernel Transport Network (kTN) such as availability, reliability, throughput and load tolerance becomes more crucial than for existing computer networks. The authors have been studying and developing a super-integrated data-driven processor to be applied to the TINA kTN nodes and network interfaces in CUE (Coordinating Users requirements and Engineering constraints) project. Since the processor is primarily designed to be a scalable VLSI component, it is easily interconnected to form a super-integrated chip and multi-chip system for achieving the performance and reliability demanded in a TINA environment. We first examine the requirements for kTN. A stream-oriented data-driven architecture is then proposed with special emphasis on effective multiprocessing capability with overload tolerance. After that, we demonstrate that autonomous load balancing among super-integrated data-driven processors without adding any runtime overhead to achieve effective and reliable multiprocessing is possible by utilizing the overload tolerance of the processor. Finally, this paper shows preliminary performance estimations of the super-integrated data-driven processor being developed to perform efficient multiprocessing in protocol handling such as TCP/IP.