Consensus-driven distributable thread scheduling in networked embedded systems
EUC'07 Proceedings of the 2007 international conference on Embedded and ubiquitous computing
| Hi-index | 0.01 |
Real time distributed transactional systems allow users to update databases which may be geographically dispersed using transactions. In addition to maintaining database consistency, these requests must terminate before their deadlines. Thus, scheduling algorithms need to enforce concurrency control and timeliness. This work studies through simulation three scheduling algorithms for real dime distributed transactional systems satisfying both serializability and timeliness. The first algorithm is based on consensus while the others use tokens to obtain globally consistent orderings. Actions are then scheduled using non-preemptive earliest deadline first. We present simulation results for each algorithm and we study the effects of some system parameters. Finally, we compare these algorithms.