Search for: All records

In this paper, we demonstrate a method of measuring the flexural elastic modulus of ceramics at an intermediate (~millimeter) scale at high temperatures. We used a picosecond laser to precisely cut microbeams from the location of interest in a bulk ceramic. They had a cross-section of approximately 100 μm × 300 μm and a length of ~1 cm. They were then tested in a thermal mechanical analyzer at room temperature, 500 °C, 800 °C, and 1100 °C using the four-point flexural testing method. We compared the elastic moduli of high-purity Al2O3 and AlN measured by our method with the reported values in the literature and found that the difference was less than 5% for both materials. This paper provides a new and accurate method of characterizing the high-temperature elastic modulus of miniature samples extracted from representative/selected areas of bulk materials. 

Accommodating long-running deep learning (DL) training and inference jobs is challenging on GPU clusters that use traditional batch schedulers, such as Slurm. Given fixed wall clock time limits, DL researchers usually need to run a sequence of batch jobs and experience long interruptions on overloaded machines. Such interruptions significantly lower the research productivity and QoS for services that are deployed in production. To mitigate the issues from interruption, we propose the design of a proactive provisioner and investigate a set of statistical learning and reinforcement learning (RL) techniques, including random forest, xgboost, Deep Q-Network, and policy gradient. Using production job traces from three GPU clusters, we train each model using a subset of the trace and then evaluate their generality using the remaining validation subset. We introduce Mirage, a Slurm-compatible resource provisioner that integrates the candidate ML methods. Our experiments show that the Mirage can reduce interruption by 17--100% and safeguard 23%-76% of jobs with zero interruption across varying load levels on the three clusters. 

Abstract Predicting and elucidating the impacts of materials on human health and the environment is an unending task that has taken on special significance in the context of nanomaterials research over the last two decades. The properties of materials in environmental and physiological media are dynamic, reflecting the complex interactions between materials and these media. This dynamic behavior requires special consideration in the design of databases and data curation that allow for subsequent comparability and interrogation of the data from potentially diverse sources. We present two data processing methods that can be integrated into the experimental process to encourage pre-mediated interoperability of disparate material data: Knowledge Mapping and Instance Mapping. Originally developed as a framework for the NanoInformatics Knowledge Commons (NIKC) database, this architecture and associated methods can be used independently of the NIKC and applied across multiple subfields of nanotechnology and material science. 

Autofluorescence (AF) poses challenges for detecting proteins of interest in situ when employing immunofluorescence (IF) microscopy. This interference is particularly pronounced in strongly autofluorescent tissues such as myocardium, where tissue AF can be comparable to IF. Although various histochemical methods have been developed to achieve effective AF suppression in different types of tissue, their applications on myocardial  samples have not been well validated. Due to inconsistency across different autofluorescent structures in sometypes of tissue, it is unclear if these methods can effectively suppress AF across all autofluorescent structures within the myocardium. Here, we quantitatively evaluated the performance of several commonly used quenching treatments on formaldehyde-fixed myocardial samples, including 0.3 M glycine, 0.3% Sudan Black B (SBB), 0.1% and 1% sodium borohydride (NaBH4), TrueVIEW® and TrueBlack®. We further assessed their quenching performance by employing the pre-treatment and post-treatment protocols, designed to cover two common IF staining scenarios where buffers contained detergents or not. The results suggest that SBB and TrueBlack® outperform other reagents in AF suppression on formaldehyde-fixed myocardial samples in both protocols. Furthermore, we inspected the quenching performance of SBB and TrueBlack® on major autofluorescent myocardial structures and evaluated their influence on IF imaging. The results suggest that SBB outperforms TrueBlack® in quenching major autofluorescent structures, while TrueBlack® excels in preserving IF labeling signal. Surprisingly, we found the treatment of NaBH4 increased AF signal and enhanced the AF contrast of major autofluorescent structures. This finding suggests that NaBH4 has the potential to act as an AF enhancer and may facilitate the interpretation of myocardial structures without the need for counterstaining. 

A burst buffer is a common method to bridge the performance gap between the I/O needs of modern supercomputing applications and the performance of the shared file system on large-scale supercomputers. However, existing I/O sharing methods require resource isolation, offline profiling, or repeated execution that significantly limit the utilization and applicability of these systems. Here we present ThemisIO, a policy-driven I/O sharing framework for a remote-shared burst buffer: a dedicated group of I/O nodes, each with a local storage device. ThemisIO preserves high utilization by implementing opportunity fairness so that it can reallocate unused I/O resources to other applications. ThemisIO accurately and efficiently allocates I/O cycles among applications, purely based on real-time I/O behavior without requiring user-supplied information or offline-profiled application characteristics. ThemisIO supports a variety of fair sharing policies, such as user-fair, size-fair, as well as composite policies, e.g., group-then-user-fair. All these features are enabled by its statistical token design. ThemisIO can alter the execution order of incoming I/O requests based on assigned tokens to precisely balance I/O cycles between applications via time slicing, thereby enforcing processing isolation. Experiments using I/O benchmarks show that ThemisIO sustains 13.5--13.7% higher I/O throughput and 19.5--40.4% lower performance variation than existing algorithms. For real applications, ThemisIO significantly reduces the slowdown by 59.1--99.8% caused by I/O interference.