Fault-tolerance has always been an important topic when it comes to running massively parallel programs at scale. Statistically, hardware and software failures are expected to occur more often on systems gathering millions of computing units. Moreover, the larger jobs are, the more computing hours would be wasted by a crash. In this paper, we describe the work done in our MPI runtime to enable both transparent and application-level checkpointing mechanisms. Unlike the MPI 4.0 User-Level Failure Mitigation (ULFM) interface, our work targets solely Checkpoint/Restart and ignores other features such as resiliency. We show how existing checkpointing methods can be practically applied to a thread-based MPI implementation given sufficient runtime collaboration. The two main contributions are the preservation of high-speed network performance during transparent C/R and the over-subscription of checkpoint data replication thanks to a dedicated user-level scheduler support. These techniques are measured on MPI benchmarks such as IMB, Lulesh and Heatdis, and associated overhead and trade-offs are discussed.

No abstract

No abstract

No abstract

The moisture has been recognized to have a significant influence on the initiation movement of sand by wind and consequently on sand transport rates. The pertinent literature regarding these phenomena is sparse and current available theoretical and empirical models exhibit considerable disagreement regarding the magnitude of moisture effects. We believe that these discrepancies comes from the fact that the moisture levels are not well controlled neither properly measured and are susceptible to strongly vary over time due to evaporation. To get rid of the variability of moisture content due to evaporation, we propose a new approach based on the use of non-volatile liquid, namely silicon oil instead of water. This insures a proper control of the liquid content and the production of reliable data concerning the variation of the transport threshold with liquid content.