Summary NERSC's newest system, Perlmutter, features a 35 PB all‐flash Lustre file system built on HPE Cray ClusterStor E1000. We present its architecture, early performance figures, and performance considerations unique to this architecture. We demonstrate the performance of E1000 OSSes through low‐level Lustre tests that achieve over 90% of the theoretical bandwidth of the SSDs at the OST and LNet levels. We also show end‐to‐end performance for both traditional dimensions of I/O performance (peak bulk‐synchronous bandwidth) and nonoptimal workloads endemic to production computing (small, incoherent I/Os at random offsets) and compare them to NERSC's previous system, Cori, to illustrate that Perlmutter achieves the performance of a burst buffer and the resilience of a scratch file system. Finally, we discuss performance considerations unique to all‐flash Lustre and present ways in which users and HPC facilities can adjust their I/O patterns and operations to make optimal use of such architectures.

Exciting changes are coming to the fore in the world of Storage and I/O for High Performance and Data Intensive computing. Existing data access and data retrieval methods that suitably support the new generation of data intensive applications in the realm of HPC and AI are being re-assessed. These new applications need to support both evolutionary as well as revolutionary approaches to data access and storage. Open source is enabling the adoption of these new techniques. This workshop looks at cutting edge trends in storage systems and solutions for data intensive HPC and AI applications with specific focus primarily on community driven initiatives and commercial products that may inspire the community.

I/O efficiency is crucial to productivity in scientific computing, but the growing complexity of HPC systems and applications complicates efforts to understand and optimize I/O behavior at scale. Data-driven machine learning-based I/O throughput models offer a solution: they can be used to identify bottlenecks, automate I/O tuning, or optimize job scheduling with minimal human intervention. Unfortunately, current state-of-the-art I/O models are not robust enough for production use and underperform after being deployed. We analyze four years of application, scheduler, and storage system logs on two leadership-class HPC platforms to understand why I/O models underperform in practice. We propose a taxonomy consisting of five categories of I/O modeling errors: poor application and system modeling, inadequate dataset coverage, I/O contention, and I/O noise. We develop litmus tests to quantify each category, allowing researchers to narrow down failure modes, enhance I/O throughput models, and improve future generations of HPC logging and analysis tools.

To avoid access to PFS, dedicated BB allocation is preferred despite of severe BB underutilization. Recently, new all-flash HPC storage systems with integrated BB and PFS are proposed, which speed up access to PFS. For this reason, we adopt BB over-subscription allocation method by allowing HPC applications to use BB only for I/O phase for improving BB utilization. Unfortunately, BB over-subscription aggravates I/O interference and demotion overhead from BB to PFS, resulting in degraded performance. To minimize the performance degradation, we develop an I/O scheduler to prevent I/O congestion and a new transparent data management system based on checkpoint/restart characteristics of HPC applications. With the proposed approach, not only the BB utilization can be improved, but also high performance of applications is achieved. In our experiments, we find that BB utilization is improved at least 2.2×, and more stable and higher checkpoint performance is guaranteed compared to other approaches. Besides, we achieve up to 96.4% hit ratio of restart requests on BB and up to 3.1× higher restart performance than others.

Scientific applications frequently spend a large fraction of their execution time in reading and writing data on parallel file systems. Identifying these I/O performance bottlenecks and attributing root causes are critical steps toward devising optimization strategies. Several existing studies analyze I/O logs of a set of benchmarks or applications that were run with controlled behaviors. However, there is still a lack of general approach that systematically identifies I/O performance bottlenecks for applications running "in the wild" on production systems. In this study, we have developed an analysis approach of "zooming in" from platform-wide to application-wide to job-level I/O logs for identifying I/O bottlenecks in arbitrary scientific applications. We analyze the logs collected on a Cray XC40 system in production over a two-month period. This study results in several insights for application developers to use in optimizing I/O behavior.

Supervised learning is a promising approach for modeling the performance of applications running on large HPC systems. A key assumption in supervised learning is that the training and testing data are obtained under the same conditions. However, in production HPC systems these conditions might not hold because the conditions of the platform can change over time as a result of hardware degradation, hardware replacement, software upgrade, and configuration updates. These changes could alter the data distribution in a way that affects the accuracy of the predictive performance models and render them less useful; this phenomenon is referred to as concept drift. Ignoring concept drift can lead to suboptimal resource usage and decreased efficiency when those performance models are deployed for tuning and job scheduling in production systems. To address this issue, we propose a concept-drift-aware predictive modeling approach that comprises two components: (1) an online Bayesian changepoint detection method that can automatically identify the location of events that lead to concept drift in near-real time and (2) a moment-matching transformation inspired by transfer learning that converts the training data collected before the drift to be useful for retraining.

Network congestion is one of the biggest problems facing HPC systems today, affecting system throughput, performance, user experience, and reproducibility. Congestion manifests as run-to-run variability due to contention for shared resources (e.g., filesystems) or routes between compute endpoints. Despite its significance, current network benchmarks fail to proxy the real-world network utilization seen on congested systems. We propose a new open-source benchmark suite called the Global Performance and Congestion Network Tests (GPCNeT) to advance the state of the practice in this area. The guiding principles used in designing GPCNeT are described and the methodology employed to maximize its utility is presented. The capabilities of GPCNeT are evaluated by analyzing results from several world's largest HPC systems, including an evaluation of congestion management on a next-generation network. The results show that systems of all technologies and scales are susceptible to congestion and this work motivates the need for congestion control in next-generation networks.

Large-scale applications typically spend a large fraction of their execution time performing I/O to a parallel storage system. However, with rapid progress in compute and storage system stack of large-scale systems, it is critical to investigate and update our understanding of the I/O behavior of large-scale applications. Toward that end, in this work, we monitor, collect and analyze a year worth of storage system data from a large-scale production parallel storage system. We perform temporal, spatial and correlative analysis of the system and uncover surprising patterns which defy existing assumptions and have important implications for future systems.

The following topics are dealt with: parallel processing; message passing; parallel machines; application program interfaces; multiprocessing systems; resource allocation; scheduling; cache storage; data analysis; graphics processing units.

Burst buffers tolerate I/O spikes in High-Performance Computing environments by using a non-volatile flash technology. Burst buffers are commonly located between parallel file systems and compute nodes, handling bursty I/Os in the middle. In this architecture, burst buffers are shared resources. The performance of an SSD is significantly reduced when it is used excessively because of garbage collection, and we have observed that SSDs in a burst buffer become slow when many users simultaneously use the burst buffer. To mitigate the performance problem, we propose a new user-level I/O isolation framework in a High-Performance Computing environment using a multi-streamed SSD. The multi-streamed SSD allocates the same flash block for I/Os in the same stream. We assign a different stream to each user; thus, the user can use the stream exclusively. To evaluate the performance, we have used open-source supercomputing workloads and I/O traces from real workloads in the Cori supercomputer at the National Energy Research Scientific Computing Center. Via user-level I/O isolation, we have obtained up to a 125% performance improvement in terms of I/O throughput. In addition, our approach reduces the write amplification in the SSDs, leading to improved SSD endurance. This user-level I/O isolation framework could be applied to deployed burst buffers without having to make any user interface changes.

MOTIVATION: Next-generation sequencing (NGS) technologies have become much more efficient, allowing whole human genomes to be sequenced faster and cheaper than ever before. However, processing the raw sequence reads associated with NGS technologies requires care and sophistication in order to draw compelling inferences about phenotypic consequences of variation in human genomes. It has been shown that different approaches to variant calling from NGS data can lead to different conclusions. Ensuring appropriate accuracy and quality in variant calling can come at a computational cost. RESULTS: We describe our experience implementing and evaluating a group-based approach to calling variants on large numbers of whole human genomes. We explore the influence of many factors that may impact the accuracy and efficiency of group-based variant calling, including group size, the biogeographical backgrounds of the individuals who have been sequenced, and the computing environment used. We make efficient use of the Gordon supercomputer cluster at the San Diego Supercomputer Center by incorporating job-packing and parallelization considerations into our workflow while calling variants on 437 whole human genomes generated as part of large association study. CONCLUSIONS: We ultimately find that our workflow resulted in high-quality variant calls in a computationally efficient manner. We argue that studies like ours should motivate further investigations combining hardware-oriented advances in computing systems with algorithmic developments to tackle emerging 'big data' problems in biomedical research brought on by the expansion of NGS technologies.