Intelligent agents for fault tolerance: from multi-agent simulation to cluster-based implementation

Recent research in multi-agent systems incorporate fault tolerance concepts, but does not explore the extension and implementation of such ideas for large scale parallel computing systems. The work reported in this paper investigates a swarm array computing approach, namely 'Intelligent Agents'. A task to be executed on a parallel computing system is decomposed to sub-tasks and mapped onto agents that traverse an abstracted hardware layer. The agents intercommunicate across processors to share information during the event of a predicted core/processor failure and for successfully completing the task. The feasibility of the approach is validated by simulations on an FPGA using a multi-agent simulator, and implementation of a parallel reduction algorithm on a computer cluster using the Message Passing Interface.

Foliar salt tolerance of Acer genotypes using chlorophyll fluorescence

The effect of increasing salinity on a range of chlorophyll fluorescence parameters in foliar tissue of 30 Acer genotypes was examined. The magnitude of the fluorescence responses differed among genotypes ranging from minor effects to substantial leaf tissue damage. Interpretation of the fluorescence expressions provided an insight into mechanisms of salt damage and resilience among genotypes. Based on reductions in a performance index (PIp) following salinity, genotypes were ranked in order from tolerant to sensitive. Based on this ranking criterion, marked differences in salt tolerance among genotypes were distinguished. It is concluded that chlorophyll fluorescence offers a rapid screening technique for assessing the foliar salinity tolerance of urban trees.

Low-temperature-induced changes in trehalose, mannitol and arabitol associated with enhanced tolerance to freezing in ectomycorrhizal basidiomycetes ( Hebeloma spp.)

Ectomycorrhizal fungi have been shown to survive sub-zero temperatures in axenic culture and in the field. However, the physiological basis for resistance to freezing is poorly understood. In order to survive freezing, mycelia must synthesise compounds that pro-tect the cells from frost damage, and certain fungal-spe-cific soluble carbohydrates have been implicated in this role. Tissue concentrations of arabitol, mannitol and trehalose were measured in axenic cultures of eight Hebeloma strains of arctic and temperate origin grown at 22, 12, 6 and 2°C. In a separate experiment, mycelia were frozen to –5°C after pre-conditioning at either 2°C or 22°C. For some, especially temperate strains, there was a clear increase in specific soluble carbohydrates at lower growth temperatures. Trehalose and mannitol were present in all strains and the highest concentrations (close to 2.5% and 0.5% dry wt.) were recorded only after a cold period. Arabitol was found in four strains only when grown at low temperature. Cold pre-condi-tioning enhanced recovery of mycelia following freez-ing. In four out of eight strains, this was paralleled by increases in mannitol and trehalose concentration at low temperature that presumably contribute towards cryopro-tection. The results are discussed in an ecological con-text with regard to mycelial overwintering in soil.

Scalable and fault tolerant failure detection and consensus

Future extreme-scale high-performance computing systems will be required to work under frequent component failures. The MPI Forum's User Level Failure Mitigation proposal has introduced an operation, MPI_Comm_shrink, to synchronize the alive processes on the list of failed processes, so that applications can continue to execute even in the presence of failures by adopting algorithm-based fault tolerance techniques. This MPI_Comm_shrink operation requires a fault tolerant failure detection and consensus algorithm. This paper presents and compares two novel failure detection and consensus algorithms. The proposed algorithms are based on Gossip protocols and are inherently fault-tolerant and scalable. The proposed algorithms were implemented and tested using the Extreme-scale Simulator. The results show that in both algorithms the number of Gossip cycles to achieve global consensus scales logarithmically with system size. The second algorithm also shows better scalability in terms of memory and network bandwidth usage and a perfect synchronization in achieving global consensus.