International Business Machines Corporation

Advanced Micro Devices Corporation

Angela Beth Dalton

William Joseph Piazza

Richard Edwin Harper

Gabriel H Loh

Michael Ignatowski

Michael J Schulte

Ganesh Suryanarayan Dasika

Valentina Salapura

Patrick M Bland

Yiyu L Chen

Thomas Michael Bradicich

Amitayu Das

An accelerator device includes a first processing unit to access a structure of a graph dataset, and a second processing unit coupled with the first processing unit to perform computations based on data values in the graph dataset.

Flexible, scalable graph-processing accelerator

Thermal diagnostic systems and methods are provided for improved detection of airflow anomalies in a computer system. In particular, processor load is selectively increased to amplify the effects caused by any airflow anomaly that may be present in the computer system. Workload migration may be used to shift processor load from another node to a target node. Artificial load may also be generated on the target node. The processor load increased to a level sufficient that an airflow anomaly would cause a detectable temperature difference at the selected node. The processor load may be increased by an amount computed to generate this detectable temperature difference. Alternatively, the processor load may be increased by a predetermined amount of between 40% and 100% of full processor utilization. While at the increased processor load, actual temperature sensed by temperature sensors may be compared to temperatures predicted from the model to detect the presence or absence of an airflow anomaly.

Detection of airflow anomalies in electronic equipment

An improved method is provided for managing workload on a multi-server computer system. In one embodiment, a subset of servers is selected according to an anticipated net workload. The remaining servers in the system may be powered off to conserve energy and prolong equipment life. Workload is dynamically apportioned among the subset of servers at selected intervals to more uniformly distribute the mean and variance of the workload among the subset of servers. More particularly, the mean and the variance for each of a plurality of workload units are equally weighed in determining a ranking of the workload units. The workload units may be ordered according to a mathematical combination of the mean and variance, such as the sum or product of mean and variance for each workload unit. The workload units are allocated among the subset of servers in according to rank, such as by assigning the workload units to the servers in a reverse round-robin fashion according to rank. Predictive power management schemes such as DVS and DVFS may then be used to control power to the servers.

Workload apportionment according to mean and variance

A system and method of detecting recirculation within a rack server system. A heat transfer model is constructed for a rack server system. A recirculation zone is specified, and hypothetical recirculation temperatures are input at the recirculation zone. The heat transfer model predicts temperatures elsewhere in the rack severe system, and a predicted temperature profile is computed. Actual temperatures in the rack server system are sensed, and an actual temperature profile is also generated. The actual temperature profile is compared with the predicted temperature profile to detect potential recirculation.

Method of detecting recirculation of heated air within a rack enclosure

Growing community of inventors

Austin, TX, United States of America

Angela Beth Dalton