Fault Tolerance on Large Scale Systems using Adaptive Process Replication

Exascale systems of the future are predicted to have mean time between failures (MTBF) of less than one hour. At such low MTBFs, employing periodic checkpointing alone will result in low efficiency because of the high number of application failures resulting in large amount of lost work due to rollbacks. In such scenarios, it is highly necessary to have proactive fault tolerance mechanisms that can help avoid significant number of failures. In this work, we have developed a mechanism for proactive fault tolerance using partial replication of a set of application processes. Our fault tolerance framework adaptively changes the set of replicated processes periodically based on failure predictions to avoid failures. We have developed an MPI prototype implementation, PAREP-MPI that allows changing the replica set. We have shown that our strategy involving adaptive process replication significantly outperforms existing mechanisms providing up to 20 percent improvement in application efficiency even for exascale systems.

In the elephant's seed shadow: the prospects of domestic bovids as replacement dispersers of three tropical Asian trees

As populations of the world's largest animal species decline, it is unclear how ecosystems will react to their local extirpation. Due to the unique ecological characteristics of megaherbivores such as elephants, seed dispersal is one ecosystem process that may be affected as populations of large animals are decimated. In typically disturbed South Asian ecosystems, domestic bovids (cattle, Bos primigenius, and buffalo, Bubalus bubalis) may often be the species most available to replace Asian elephants (Elephas maximus) as endozoochorous dispersers of large-fruited mammal-dispersed species. We use feeding trials, germination trials, and movement data from the tropical moist forests of Buxa Tiger Reserve (India) to examine whether domestic bovids are viable replacements for elephants in the dispersal of three largefruited species: Dillenia indica, Artocarpus chaplasha, and Careya arborea. We find that (1) once consumed, seeds are between 2.5 (C. arborea) and 26.5 (D. indica) times more likely to pass undigested into elephant dung than domestic bovid dung; and (2) seeds from elephant dung germinated as well as or better than seeds taken from bovid dung for all plant species, with D. indica seeds from elephant dung 1.5 times more likely to germinate. Furthermore, since wild elephants have less constrained movements than even free-roaming domestic bovids, we calculate that maximum dispersal by elephants is between 9.5 and 11.2 times farther than that of domestic bovids, with about 20% of elephant-dispersed seeds being moved farther than the maximum distance seeds are moved by bovids. Our findings suggest that, while bovids are able to disperse substantial numbers of seeds over moderate distances for two of the three study species, domestic bovids will be unable to routinely emulate the reliable, long-distance dispersal of seeds executed by elephants in this tropical moist forest. Thus while domestic bovids can attenuate the effects of losing elephants as dispersers, they may not be able to prevent the decline of various mammal-dispersed fruiting species in the face of overhunting, habitat fragmentation, and climate change.

Fruit Trait Preference in Rhesus Macaques (Macaca mulatta) and its Implications for Seed Dispersal

Primates constitute 25-40 % of the frugivore biomass of tropical forests. Primate fruit preference, as a determinant of seed dispersal, can therefore have a significant impact on these ecosystems. Although the traits of fruits included in primate diets have been described, fruit trait preference has been less studied with respect to fruit availability. We examined fruit trait preference and its implications for seed dispersal in the rhesus macaque (Macaca mulatta), a dietarily flexible species and important seed disperser, at the Buxa Tiger Reserve, India. Over a year, we monitored the phenology of selected trees in the study area, observed the feeding behavior of rhesus macaques using scans and focal animal sampling, and documented morphological traits of the fruits/seeds consumed. Using generalized linear modeling, we found that the kind of edible tissue was the chief determinant of fruit consumption, with M. mulatta feeding primarily on fruits with juicy-soft pulp and acting as seed predators for those with no discernible pulp. Overall, the preferred traits were external covers that could be easily pierced by a fingernail, medium to large seeds, true stone-like seeds, and juicy-soft edible tissue, thereby implying that fruit taxa with these traits had a higher probability of being dispersed. Macaques were more selective during the high fruit availability period than the low fruit availability period, preferentially feeding on soft-skinned fruits with juicy-soft pulp. We suggest that further studies be conducted across habitats and time to understand the consistency of interactions between primates and fruits with specific traits to determine the degree of selective pressure (if any) that is exerted by primates on fruit traits.

Infiltration of Matrix-Non-producers Weakens the Salmonella Biofilm and Impairs Its Antimicrobial Tolerance and Pathogenicity

Bacterial biofilms display a collective lifestyle, wherein the cells secrete extracellular polymeric substances (EPS) that helps in adhesion, aggregation, stability, and to protect the bacteria from antimicrobials. We asked whether the BPS could act as a public good for the biofilm and observed that infiltration of cells that do not produce matrix components weakened the biofilm of Salmonella enterica serovar Typhimurium. PS production was costly for the producing cells, as indicated by a significant reduction in the fitness of wild type (WT) cells during competitive planktonic growth relative to the non-producers. Infiltration frequency of non-producers in the biofilm showed a concomitant decrease in overall productivity. It was apparent in the confocal images that the non producing cells benefit from the BPS produced by the Wild Type (WT) to stay in the biofilm. The biofilm containing non-producing cells were more significantly susceptible to sodium hypochlorite and ciprofloxacin treatment than the WT biofilm. Biofilm infiltrated with non-producers delayed the pathogenesis, as tested in a murine model. The cell types were spatially assorted, with non producers being edged out in the biofilm. However, cellulose was found to act as a barrier to keep the non-producers away from the WT microcolony. Our results show that the infiltration of non-cooperating cell types can substantially weaken the biofilm making it vulnerable to antibacterials and delay their pathogenesis. Cellulose, a component of BPS, was shown to play a pivotal role of acting as the main public good, and to edge-out the non-producers away from the cooperating microcolony.

How much Dillenia indica seed predation occurs from Asian elephant dung?

Elephants are thought to be effective seed dispersers, but research on whether elephant dung effectively protects seeds from seed predation is lacking. Quantifying rates of seed predation from elephant dung will facilitate comparisons between elephants and alternative dispersers, helping us understand the functional role of megaherbivores in ecosystems. We conducted an experiment to quantify the predation of Dillenia indica seeds from elephant dung in Buxa Reserve, India from December 2012 to April 2013. Using dung boluses from the same dung pile, we compared the number of seeds in boluses that are a) opened immediately upon detection (control boluses), b) made available only to small seed predators (<= 3 mm wide) for 1-4 months, and c) made available to all seed predators and secondary dispersers for 1-4 months. Using a model built on this experiment, we estimated that seed predation by small seed predators (most likely ants and termites) destroys between 82.9% and 96.4% of seeds in elephant dung between the time of defecation and the median germination date for D. indica. Exposure to larger seed predators and secondary dispersers did not lead to a significant additional reduction in the number of seeds per dung bolus. Our findings suggest that post-dispersal seed predation by small insects (<3 mm) substantially reduces but does not eliminate the success of elephants as dispersers of D. indica in a tropical moist forest habitat. (C) 2015 Elsevier Masson SAS. All rights reserved.