Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “WELL LOGGING”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 811 records · Page 45

DISCOVERY OF PULSATIONS, INCLUDING POSSIBLE PRESSURE MODES, IN TWO NEW EXTREMELY LOW MASS, He-CORE WHITE DWARFS

We report the discovery of the second and third pulsating extremely low mass (ELM) white dwarfs (WDs), SDSS J111215.82+111745.0 (hereafter J1112) and SDSS J151826.68+065813.2 (hereafter J1518). Both have masses < 0.25 M{sub Sun} and effective temperatures below 10, 000 K, establishing these putatively He-core WDs as a cooler class of pulsating hydrogen-atmosphere WDs (DAVs, or ZZ Ceti stars). The short-period pulsations evidenced in the light curve of J1112 may also represent the first observation of acoustic (p-mode) pulsations in any WD, which provide an exciting opportunity to probe this WD in a complimentary way compared to the long-period g-modes that are also present. J1112 is a T{sub eff} =9590 {+-} 140 K and log g =6.36 {+-} 0.06 WD. The star displays sinusoidal variability at five distinct periodicities between 1792 and 2855 s. In this star, we also see short-period variability, strongest at 134.3 s, well short of the expected g-modes for such a low-mass WD. The other new pulsating WD, J1518, is a T{sub eff} =9900 {+-} 140 K and log g =6.80 {+-} 0.05 WD. The light curve of J1518 is highly non-sinusoidal, with at least seven significant periods between 1335 and 3848 s. Consistent with the expectation that ELM WDs must be formed in binaries, these two new pulsating He-core WDs, in addition to the prototype SDSS J184037.78+642312.3, have close companions. However, the observed variability is inconsistent with tidally induced pulsations and is so far best explained by the same hydrogen partial-ionization driving mechanism at work in classic C/O-core ZZ Ceti stars.

79 ASTRONOMY AND ASTROPHYSICS↗

The SRG/eROSITA All-Sky Survey: Optical identification and properties of galaxy clusters and groups in the western galactic hemisphere

The first SRG/eROSITA All-Sky Survey (eRASS1) provides the largest intracluster medium-selected galaxy cluster and group catalog covering the western Galactic hemisphere. Compared to samples selected purely on X-ray extent, the sample purity can be enhanced by identifying cluster candidates using optical and near-infrared data from the DESI Legacy Imaging Surveys. Using the red-sequence-based cluster findereROMaPPer, we measured individual photometric properties (redshiftz λ , richnessλ, optical center, and BCG position) for 12000 eRASS1 clusters over a sky area of 13 116 deg 2 , augmented by 247 cases identified by matching the candidates with known clusters from the literature. The median redshift of the identified eRASS1 sample isz= 0.31, with 10% of the clusters atz> 0.72. The photometric redshifts have an accuracy ofδz/(1 +z) ≲ 0.005 for 0.05 specand velocity dispersionσ) were measured a posteriori for a subsample of 3210 and 1499 eRASS1 clusters, respectively, using an extensive compilation of spectroscopic redshifts of galaxies from the literature. We infer that the primary eRASS1 sample has a purity of 86% and optical completeness >95% forz> 0.05. For these and further quality assessments of the eRASS1 identified catalog, we applied our identification method to a collection of galaxy cluster catalogs in the literature, as well as blindly on the full Legacy Surveys covering 24069 deg 2 . Using a combination of these cluster samples, we investigated the velocity dispersion-richness relation, finding that it scales with richness as log(λ norm ) = 2.401 × log(σ) − 5.074 with an intrinsic scatter ofδ in = 0.10 ± 0.01 dex. The primary product of our work is the identified eRASS1 cluster catalog with high purity and a well-defined X-ray selection process, opening the path for precise cosmological analyses presented in companion papers.

Astronomy & Astrophysics↗

Linking Stream Chemistry to Subsurface Redox Architecture

As drinking-water scarcity grows worldwide, we need to improve predictions of the quantity and quality of our water resources. An overarching problem for model improvement is that we do not know the geological structure of aquifers in sufficient detail. In this work, we demonstrate that mineral-water reactions imprint structure in the subsurface that impacts the flow and transport of some chemical species. Specifically, pyrite, a ubiquitous mineral, commonly oxidizes and depletes in the upper layers of the weathering profile in most humid watersheds, only remaining at depths of meters. We hypothesize that variations in concentrations (C) of pyrite-derived sulfate released into rivers as a function of discharge (q) reflect the rate-limiting step and depth of pyrite-oxidizing layers. We found that log C - log q behaviors thus differ in small and large watersheds in the Susquehanna River Basin as well as in selected watersheds in the Western United States. Although coal mining changes pyrite oxidation from closed to open system with respect to O 2 , patterns in stream chemistry as a function of discharge are consistent with deep and shallow pyrite oxidation zones in small and large watersheds respectively. Therefore, understanding the subsurface patterns of mineral reactions and how they affect the architecture of aquifers will elucidate patterns of changing river chemistry and our ability to manage water resources in the future under accelerated land use and climate change.

54 ENVIRONMENTAL SCIENCES↗

Molecular mobility in glassy dispersions

Dielectric spectroscopy was used to characterize the structural relaxation in pharmaceutical dispersions containing nifedipine (NIF) and either poly(vinyl) pyrrolidone (PVP) or hydroxypropyl methylcellulose acetate succinate (HPMCAS). The shape of the dielectric response (permittivity versus log time) curve was observed to be independent of temperature. Thus, for the pure NIF as well as the dispersions, the validity of the time-temperature superposition principle was established. Furthermore, though the shape of the full dielectric response varied with polymer concentration, the regime related to the α- or structural relaxation was found to superimpose for the dispersions, though not with the response of the NIF itself. Hence, there is a limited time-temperature-concentration superposition for these systems as well. Therefore, in this polymer concentration range, calculation of long relaxation times in these glass-forming systems becomes possible. We found that strong drug-polymer hydrogen bonding interactions improved the physical stability (i.e., delayed crystallization) by reducing the molecular mobility. The strength of hydrogen bonding, structural relaxation time, and crystallization followed the order: NIF$-$PV P>NIF$-$HPMCAS>NIF. With an increase in polymer concentration, the relaxation times were longer indicating a decrease in molecular mobility. The temperature dependence of relaxation time, in other words fragility, was independent of polymer concentration. This is the first application of the superposition principle to characterize structural relaxation in glassy pharmaceutical dispersions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

HATS-3b: AN INFLATED HOT JUPITER TRANSITING AN F-TYPE STAR

We report the discovery by the HATSouth survey of HATS-3b, a transiting extrasolar planet orbiting a V = 12.4 F dwarf star. HATS-3b has a period of P = 3.5479 days, mass of M{sub p} = 1.07 M {sub J}, and radius of R{sub p} = 1.38 R {sub J}. Given the radius of the planet, the brightness of the host star, and the stellar rotational velocity (vsin i = 9.0 km s{sup –1}), this system will make an interesting target for future observations to measure the Rossiter-McLaughlin effect and determine its spin-orbit alignment. We detail the low-/medium-resolution reconnaissance spectroscopy that we are now using to deal with large numbers of transiting planet candidates produced by the HATSouth survey. We show that this important step in discovering planets produces log g and T {sub eff} parameters at a precision suitable for efficient candidate vetting, as well as efficiently identifying stellar mass eclipsing binaries with radial velocity semi-amplitudes as low as 1 km s{sup –1}.

79 ASTRONOMY AND ASTROPHYSICS↗

EXTRAGALACTIC CHEMICAL ABUNDANCES: DO H II REGIONS AND YOUNG STARS TELL THE SAME STORY? THE CASE OF THE SPIRAL GALAXY NGC 300

We have obtained new spectrophotometric data for 28 H II regions in the spiral galaxy NGC 300, a member of the nearby Sculptor Group. The detection of several auroral lines, including [O III] {lambda}4363, [S III] {lambda}6312, and [N II] {lambda}5755, has allowed us to measure electron temperatures and direct chemical abundances for the whole sample. We determine for the first time in this galaxy a radial gas-phase oxygen abundance gradient based solely on auroral lines, and obtain the following least-square solution: 12 + log(O/H) = 8.57({+-}0.02) - 0.41({+-}0.03)R/R {sub 25}, where the galactocentric distance is expressed in terms of the isophotal radius R {sub 25}. The characteristic oxygen abundance, measured at 0.4 x R{sub 25}, is 12 + log(O/H) = 8.41. The gradient corresponds to -0.077 {+-} 0.006 dex kpc{sup -1}, and agrees very well with the galactocentric trend in metallicity obtained for 29 B and A supergiants in the same galaxy, -0.081 {+-} 0.011 dex kpc{sup -1}. The intercept of the regression for the nebular data virtually coincides with the intercept obtained from the stellar data, which is 8.59({+-}0.05). This allows little room for depletion of nebular oxygen onto dust grains, although in this kind of comparison we are somewhat limited by systematic uncertainties, such as those related to the atomic parameters used to derive the chemical compositions. We discuss the implications of our result with regard to strong-line abundance indicators commonly used to estimate the chemical compositions of star-forming galaxies, such as R {sub 23}. By applying a few popular calibrations of these indices based on grids of photoionization models on the NGC 300 H II region fluxes, we find metallicities that are higher by 0.3 dex (a factor of 2) or more relative to our nebular (T{sub e} based) and stellar ones. We detect Wolf-Rayet stellar emission features in {approx}1/3 of our H II region spectra, and find that in one of the nebulae hosting these hot stars the ionizing field has a particularly hard spectrum, as gauged by the 'softness' parameter {eta} = (O{sup +}/O{sup ++})/(S{sup +}/S{sup ++}). We suggest that this is related to the presence of an early WN star. By considering a larger sample of extragalactic H II regions we confirm, using direct abundance measurements, previous findings of a metallicity dependence of {eta}, in the sense that softer stellar continua are found at high metallicity.

99 GENERAL AND MISCELLANEOUS↗

A SECOND-ORDER BIAS MODEL FOR THE LOGARITHMIC HALO MASS DENSITY

We present an analytic model for the local bias of dark matter halos in a {Lambda}CDM universe. The model uses the halo mass density instead of the halo number density and is searched for various halo mass cuts, smoothing lengths, and redshift epochs. We find that, when the logarithmic density is used, the second-order polynomial can fit the numerical relation between the halo mass distribution and the underlying matter distribution extremely well. In this model, the logarithm of the dark matter density is expanded in terms of log halo mass density to the second order. The model remains excellent for all halo mass cuts (from M{sub cut} = 3 Multiplication-Sign 10{sup 11} to 3 Multiplication-Sign 10{sup 12} h{sup -1} M{sub Sun }), smoothing scales (from R = 5 h{sup -1} Mpc to 50 h{sup -1} Mpc), and redshift ranges (from z = 0 to 1.0) considered in this study. The stochastic term in the relation is found to be not entirely random, but a part of the term can be determined by the magnitude of the shear tensor.

79 ASTRONOMY AND ASTROPHYSICS↗

Handbook of Best Practices for Geothermal Drilling

This Handbook is a description of the complex process that comprises drilling a geothermal well. The focus of the detailed Chapters covering various aspects of the process (casing design, cementing, logging and instrumentation, etc) is on techniques and hardware that have proven successful in geothermal reservoirs around the world. The Handbook will eventually be linked to the GIA web site, with the hope and expectation that it can be continually updated as new methods are demonstrated or proven.

15 GEOTHERMAL ENERGY↗

The Anatomy of an Unusual Edge-on Protoplanetary Disk. I. Dust Settling in a Cold Disk

As the earliest stage of planet formation, massive, optically thick, and gas-rich protoplanetary disks provide key insights into the physics of star and planet formation. When viewed edge-on, high-resolution images offer a unique opportunity to study both the radial and vertical structures of these disks and relate this to vertical settling, radial drift, grain growth, and changes in the midplane temperatures. In this work, we present multi-epoch Hubble Space Telescope and Keck scattered light images, and an Atacama Large Millimeter/submillimeter Array 1.3 mm continuum map for the remarkably flat edge-on protoplanetary disk SSTC2DJ163131.2–242627, a young solar-type star in ρ Ophiuchus. We model the 0.8 μm and 1.3 mm images in separate Markov Chain Monte Carlo (MCMC) runs to investigate the geometry and dust properties of the disk using the MCFOST radiative transfer code. In scattered light, we are sensitive to the smaller dust grains in the surface layers of the disk, while the submillimeter dust continuum observations probe larger grains closer to the disk midplane. An MCMC run combining both data sets using a covariance-based log-likelihood estimation was marginally successful, implying insufficient complexity in our disk model. The disk is well characterized by a flared disk model with an exponentially tapered outer edge viewed nearly edge-on, though some degree of dust settling is required to reproduce the vertically thin profile and lack of apparent flaring. A colder than expected disk midplane, evidence for dust settling, and residual radial substructures all point to a more complex radial density profile to be probed with future, higher-resolution observations.

47 OTHER INSTRUMENTATION↗

Physical and chemical properties of the regional mixed layer of Mexico's Megapolis – Part II: Evaluation of measured and modeled trace gases and particle size distributions

This study extends the work of Baumgardner et al. (2009) in which measurements of trace gases and particles at a remote, high-altitude mountain site 60 km from Mexico City were analyzed with respect to the origin of air masses. In the current evaluation, the temperature, water vapor, ozone (O3), carbon monoxide (CO), acyl peroxy nitrate (APN) and particle size distributions (PSDs) of the mass concentrations of sulfate, nitrate, ammonium and organic mass (OM) were simulated with the WRF-Chem chemical transport model and compared with the measurements at the mountain site. The model prediction of the diurnal trends of the gases were well correlated with the measurements before the regional boundary layer reached the measurement site but underestimated the concentrations after that time. The differences are caused by an overly rapid growth of the boundary layer by the model with too much dilution. There also appears to be more O3 produced by photochemical production, downwind of the emission sources, than predicted by the model. The measured and modeled PSDs compare very well with respect to their general shape and diameter of the peak concentrations. The spectra are log normally distributed with most of the mass in the accumulation mode and the geometric diameter centered at 200 ±20 nm, with little observed or predicted change with respect to the origin of the air mass or the time when the RBL is above the Altzomoni research. Only the total mass changed with time and air mass origin. The invariability of the average diameter of the accumulation mode suggests that there is very little growth of the particles by condensation or coagulation after six hours of aging downwind of the major sources of anthropogenic emissions in Mexico’s Megapolis.

Ochoa, Carlos↗

Viscosity of liquid anorthite.

The viscosity of liquid anorthite has been determined over the temperature ranges between 1450 to 1620 C and 820 to 950 C. The high-temperature data agree well with previous experimental data and with predictions of the Bottinga and Weill model. The overall log (viscosity) versus 1/T relation exhibits pronounced and rather continuous curvature. The viscosity of anorthite is higher at any given temperature and more strongly temperature-dependent than that of the anorthite-rich lunar compositions 14259, 14310, and 15418. The room-temperature density of glassy anorthite (2.64 gm/cu cm) and the thermal expansion coefficients of glassy and liquid anorthite have also been determined. The volume expansion coefficient for the glass is about 0.000018 per deg C, and that for the liquid is about 0.000048 per deg C. These values are used to relate the high-temperature flow data to the predictions of free-volume theories.

Cukierman, M.↗

Tests of the Tully-Fisher relation. 1: Scatter in infrared magnitude versus 21 cm width

We examine the precision of the Tully-Fisher relation (TFR) using a sample of galaxies in the Coma region of the sky, and find that it is good to 5% or better in measuring relative distances. Total magnitudes and disk axis ratios are derived from H and I band surface photometry, and Arecibo 21 cm profiles define the rotation speeds of the galaxies. Using 25 galaxies for which the disk inclination and 21 cm width are well defined, we find an rms deviation of 0.10 mag from a linear TFR with dI/d(log W(sub c)) = -5.6. Each galaxy is assumed to be at a distance proportional to its redshift, and an extinction correction of 1.4(1-b/a) mag is applied to the total I magnitude. The measured scatter is less than 0.15 mag using milder extinction laws from the literature. The I band TFR scatter is consistent with measurement error, and the 95% CL limits on the intrinsic scatter are 0-0.10 mag. The rms scatter using H band magnitudes is 0.20 mag (N = 17). The low width galaxies have scatter in H significantly in excess of known measurement error, but the higher width half of the galaxies have scatter consistent with measurement error. The H band TFR slope may be as steep as the I band slope. As the first applications of this tight correlation, we note the following: (1) the data for the particular spirals commonly used to define the TFR distance to the Coma cluster are inconsistent with being at a common distance and are in fact in free Hubble expansion, with an upper limit of 300 km/s on the rms peculiar line-of-sight velocity of these gas-rich spirals; and (2) the gravitational potential in the disks of these galaxies has typical ellipticity less than 5%. The published data for three nearby spiral galaxies with Cepheid distance determinations are inconsistent with our Coma TFR, suggesting that these local calibrators are either ill-measured or peculiar relative to the Coma Supercluster spirals, or that the TFR has a varying form in different locales.

Bernstein, Gary M.↗

Kalman Filter Chemical Data Assimilation: A Case Study in January 1992

This paper describes a Kalman filter chemical data assimilation system and its use for analysing a vertical atmospheric profile during January 1992. The vertical profile was at an equivalent PV latitude (phi(sub e)) of 55 deg S and consisted of 21 potential temperature (theta) levels spaced equally in log(theta) between 400 K and 2000 K. This equivalent latitude was chosen as it was well observed during January 1992 by instruments on board the Upper Atmosphere Research Satellite (UARS).

Lary, D. J.↗

Massively parallel algorithms for trace-driven cache simulations

Trace driven cache simulation is central to computer design. A trace is a very long sequence of reference lines from main memory. At the t(exp th) instant, reference x sub t is hashed into a set of cache locations, the contents of which are then compared with x sub t. If at the t sup th instant x sub t is not present in the cache, then it is said to be a miss, and is loaded into the cache set, possibly forcing the replacement of some other memory line, and making x sub t present for the (t+1) sup st instant. The problem of parallel simulation of a subtrace of N references directed to a C line cache set is considered, with the aim of determining which references are misses and related statistics. A simulation method is presented for the Least Recently Used (LRU) policy, which regradless of the set size C runs in time O(log N) using N processors on the exclusive read, exclusive write (EREW) parallel model. A simpler LRU simulation algorithm is given that runs in O(C log N) time using N/log N processors. Timings are presented of the second algorithm's implementation on the MasPar MP-1, a machine with 16384 processors. A broad class of reference based line replacement policies are considered, which includes LRU as well as the Least Frequently Used and Random replacement policies. A simulation method is presented for any such policy that on any trace of length N directed to a C line set runs in the O(C log N) time with high probability using N processors on the EREW model. The algorithms are simple, have very little space overhead, and are well suited for SIMD implementation.

Nicol, David M.↗

Breakthroughs in Seismic and Borehole Characterization of Basalt Sequestration Targets

Mafic continental flood basalts form a globally important, but under-characterized CO2 sequestration target. The Columbia River Basalt Group (CRBG) in the northwestern U.S. is up to 5 km thick and covers over 168,000 km2. In India, flood basalts are 3 km thick and cover greater than 500,000 km2. Laboratory experiments demonstrate that the CRBG and other basalts react with formation water and super critical (sc) CO2 to precipitate carbonates, thus adding a potential mineral trapping mechanism to the standard trapping mechanisms of most other types of CO2 sequestration reservoirs. Brecciated tops of individual basalt flows in the CRBG form regional aquifers that locally have greater than 30% porosity and three Darcies of permeability. Porous flow tops are potential sites for sequestration of gigatons of scCO2 in areas where the basalts contain unpotable water and are at depths greater than 800 m. In this paper we report on the U.S. DOE Big Sky Regional Carbon Sequestration Partnership surface seismic and borehole geophysical characterization that supports a field test of capacity, integrity, and geochemical reactivity of CRBG reservoirs in eastern Washington, U.S.A. Traditional surface seismic methods have had little success in imaging basalt features in on-shore areas where the basalt is thinly covered by sediment. Processing of the experimental 6.5 km, 5 line 3C seismic swath included constructing an elastic wavefield model, identifying and separating seismic wave modes, and processing the swath as a single 2D line. Important findings include: (1) a wide variety of shear wave energy modes swamp the P-wave seismic records; (2) except at very short geophone offsets, ground roll overprints P-wave signal; and (3) because of extreme velocity contrasts, P-wave events are refracted at incidence angles greater than 7-15 degrees. Subsequent removal of S-wave and other noise during processing resulted in tremendous improvement in image quality. The application of wireline logging to onshore basalts is underexploited. Full waveform sonic logs and resistivity-based image logs acquired in the 1250 m basalt pilot borehole provide powerful tools for evaluating geomechanics and lithofacies. The azimuth of the fast shear wave is parallel to SH and records the changes through time in basalt flow and tectonic stress tensors. Combined with image log data, azimuthal S-wave data provide a borehole technique for assessing basalt emplacement and cooling history that is related to the development of reservoirs and seals, as well as the orientation of tectonic stresses and fracture systems that could affect CO2 transport or containment. Reservoir and seal properties are controlled by basalt lithofacies, and rescaled P- and S- wave slowness curves, integrated with image logs, provide a tool for improved recognition of subsurface lithofacies.

54 ENVIRONMENTAL SCIENCES↗

Appendix for the Final Technical Report - DE FE0009284

Carbon capture utilization storage (CCUS) is a potential technology to store anthropogenic emissions for CO 2 . Utilization often refers to CO 2 -enhanced oil recovery (CO 2 -EOR). An important factor in the success of CO 2 storage in saline formations or CO 2 -EOR reservoirs is ensuring that the storage occurs safely and is long-term. Assessment of well integrity has become more important for CCS and CO 2 -EOR as it has become apparent that wells represent the most likely migration pathway for CO 2 to leave a CCS storage unit or a CO 2 -EOR reservoir. Although wells represent a migration pathway they are also the best vehicle for employing technology monitoring CO 2 injection and storage. This contradiction of being a potential migration path and key monitoring technology leads to a need to understand how monitoring wells may be similar or different in comparison to other types of wells with respect to migration risk. The maturation and completion US Department of Energy sponsored research projects presents an opportunity to assess the integrity of monitoring wells that have been exposed to injected CO 2 . This paper discusses an integrity assessment of two monitoring wells in an operating CO 2 -EOR flood in Mississippi, USA. The CFU31F-2 and CFU31F-3 monitoring wells were constructed to test monitoring technologies in and above a commercial CO 2 -EOR project. The materials selected and the design of the well were optimized for monitoring. Carbonation in CFU31-F2 was seen as high as 7900 ft, above what was considered top of cement based on the logs. Time-lapse comparison of cement bond amplitude data and acoustic impedance maps show a deterioration of signal that implies a deterioration of cement bond or cement along much of the cemented annulus in the long-string section. Analysis of sidewall cores using XRD and LA-ICP-MS validated the log interpretation by confirming the degradation of cement (carbonation) along the casing-cement interface. The ultrasonic image maps also clearly identify the control lines and monitoring technology attached to the outside of the of the long-string casing on each well studied. The control lines appear as microdebonded or fluid filled vertical features implying that they could act as leakage pathways. The sidewall core through the control line at 10380 ft confirms that CO 2 is migrating along the control line with heavily carbonated cement at the control line interface. LA-ICP-MS and XRD on formation interface of the sidewall cores collected in both wells indicates that CO 2 is also moving of the cement-formation interface. LA-ICP-MS and XRD indicate that the amount carbonation in the center of the cores was less than the carbonation at the interfaces. Indicating that CO 2 is reaching the center of the cores by diffusing in from the interfaces and not migrating up from the reservoir though the porous matrix of the cement. This agrees with Duguid et al. [5] and Carey et al. [1] who have found that the interfaces in the well are more conductive than the porous network of the cement. Both the materials used to construct the well and the decision to attach monitoring technology to the outside of the well may have contributed to the migration of CO 2 along the interfaces. Careful consideration should be given to material selection to ensure that it does not degrade when in contact with the fluids in the reservoir and overlying strata. The addition of the control line on the outside of the casing complicated the cement placement and likely caused no cement to bond to the casing adjacent to the control line leading out of the reservoir. Further study of other wells with external lines should be conducted to see of the results of the construction of CFU31-F2 and -F3 is normal or an exception.

54 ENVIRONMENTAL SCIENCES↗

Characterization and identification of HPC applications at leadership computing facility

High Performance Computing (HPC) is an important method for scientific discovery via large-scale simulation, data analysis, or artificial intelligence. Leadership-class supercomputers are expensive, but essential to run large HPC applications. The Petascale era of supercomputers began in 2008, with the first machines achieving performance in excess of one petaflops, and with the advent of new supercomputers in 2021 (e.g., Aurora, Frontier), the Exascale era will soon begin. However, the high theoretical computing capability (i.e., peak FLOPS) of a machine is not the only meaningful target when designing a supercomputer, as the resources demand of applications varies. A deep understanding of the characterization of applications that run on a leadership supercomputer is one of the most important ways for planning its design, development and operation. In order to improve our understanding of HPC applications, user demands and resource usage characteristics, we perform correlative analysis of various logs for different subsystems of a leadership supercomputer. This analysis reveals surprising, sometimes counter-intuitive patterns, which, in some cases, conflicts with existing assumptions, and have important implications for future system designs as well as supercomputer operations. For example, our analysis shows that while the applications spend significant time on MPI, most applications spend very little time on file I/O. Combined analysis of hardware event logs and task failure logs show that the probability of a hardware FATAL event causing task failure is low. Combined analysis of control system logs and file I/O logs reveals that pure POSIX I/O is used more widely than higher level parallel I/O. Based on holistic insights of the application gained through combined and co-analysis of multiple logs from different perspectives and general intuition, we engineer features to "fingerprint" HPC applications. We use t-SNE (a machine learning technique for dimensionality reduction) to validate the explainability of our features and finally train machine learning models to identify HPC applications or group those with similar characteristic. To the best of our knowledge, this is the first work that combines logs on file I/O, computing, and inter-node communication for insightful analysis of HPC applications in production.

Liu, Zhengchun↗