As high performance computing (HPC) systems continue to increase in size, reliable and scalable computational models become critical. Tree-based overlay networks (TBONs) help to address scalability by providing scalable data multicast, data gather, and data aggregation services. In this dissertation, we address the reliability challenges of TBON based HPC tools and applications.
We exploit the characteristics of many TBON computations to develop a new failure recovery model, state compensation, that uses: inherently redundant information from processes that survive failures to compen. sate for information lost due to failures. weak data consistency to relax the constraints of the recovery mechanisms, and o protocols that allow processes to recover from failures independently.
State compensation requires no additional computational, network or storage resources in the absence of failures. When failures do occur, a small subset of TBON processes participate in failure recovery, so failure recovery is scalable.
(Kode: UMI Number: 3348717) Rak Penelitian 2