The Tuple Space: An Old Solution to a New Problem?
Adrian Friday, Stephen P. Wade, Nigel Davies, Gordon S. Blair · Lancaster EPrints (Lancaster University) · 1997
This paper discusses the impact on current distributed systems thinking of emerging and future networking technologies. In particular, we believe that future distributed systems platforms will be forced to abandon traditional thinking and embrace a new asynchronous model to justify their existence in the future. The world's internetworks are already beginning to be affected by the rapid development of two highly significant areas of research interest: mobile computing and high speed networking. Mobile networks offer one of the most diverse and challenging environments in which to build distributed systems. A typical network is characterised by unpredictable connectivity, increased levels of bit error rate, endto -end delay and jitter, dramatically reduced bandwidth and significantly, increased cost to the end-user. Furthermore, the diversity of bearer services offer applications little consistency in terms of service, for instance, style of interaction (connectionless versus connection-oriented), error protection and group support. Examples of such services include GSM, DECT and the emerging TETRA standard. High speed fibre based networks (such as FDDI and ATM) are being deployed in response to the escalating demand for capacity driven by the rapid increase in the popularity of the internet. Future gigabit networks (currently under development) will offer transfer speeds in excess of many existing internal bus architectures [Boden,95]. The deployment of such networks in the near future will raise a great many issues. For instance, how to minimise the number of end-to-end delays incurred in interactions (latency reduction), how to utilise such a highly parallel network (distributed processing) and how to achieve scalability (particularly with respect to the predicted dem...