Abstract: A method of simulating a design described in HDL is provided. In this method, modules of the design can be partitioned into first modules for simulation by a serial simulation engine and second modules for simulation by a concurrent simulation engine. The first and second modules can be prioritized for simulation based on classes of events consistent with an execution model of the HDL. Simulations of the serial and concurrent simulation engines can be synchronized for each class of events. Synchronizing can include transferring updated interface variable values, which are shared by the second modules and at least a subset of the first modules, between the serial simulation engine and the concurrent simulation engine. This transferring can include translating representations of the updated interface variable values.

Abstract: A system that generates a model for an electronic prognostics system which is used to monitor a computer system while the computer system is operating correctly. During operation, the system receives telemetry signals generated by the computer system. Next, the system builds the model based on the telemetry signals, wherein the model uses correlations between different telemetry signals to identify anomalies in the telemetry signals. The system then validates the model against telemetry signals from the computer system.

Abstract: A system that balances thermal variations within a set of computer systems in a datacenter. During operation, the system obtains a thermal flux map for the set of computer systems. The system then analyzes the thermal flux map to determine whether imbalances exist in the thermal flux across the set of computer systems. If so, the system can adjust: (1) the scheduling of loads across the set of computer systems, and/or (2) air conditioning within the datacenter, so that the thermal flux is more balanced across the set of computer systems.

Abstract: A computer system to schedule loads across a set of processor cores is described. During operation, the computer system receives a process to be executed. Next, the computer system obtains one or more thermodynamic process characteristics associated with the process and one or more thermodynamic processor-core characteristics associated with operation of the set of processor cores. Then, the computer system schedules the process to be executed by at least one of the processor cores based on the one or more thermodynamic process characteristics and the one or more thermodynamic processor-core characteristics.

Abstract: Some embodiments of the present invention provide a system that stores telemetry data from a computer system. The system includes a first buffer, a second buffer, and a third buffer. During operation, the system periodically obtains the telemetry data from the computer system and stores the telemetry data in the first buffer, second buffer, and third buffer. The system also compresses the telemetry data in the first and second buffers. To compress the data, the system creates a first set of summary statistics from the telemetry data in the first buffer and the second buffer and stores the first set of summary statistics in the first buffer, which becomes a historical data buffer.

Abstract: Simulating hardware includes generating a data flow representation of the hardware, based on a hardware description language (HDL) description. The data flow representation including compatibility information that preserves behavioral and synthesizable characteristics of the HDL description. Simulating hardware further includes generating code from the data flow representation, and executing the code concurrently.

Abstract: Some embodiments of the present invention provide a system that characterizes a computer system using a pattern-recognition model. First, values for an environmental parameter are monitored from a set of sensors associated with the computer system. Then, a baseline for the environmental parameter is calculated based on the monitored values from a subset of the set of sensors. Next, the baseline is subtracted from the monitored values from sensors in the set of sensors to produce compensated values. Then, the compensated values are used as inputs to the pattern-recognition model, which produces estimates for the compensated values based on correlations between the compensated values learned during a training phase. Next, residuals are calculated by subtracting the estimates for the compensated values from the compensated values. Then, the residuals are analyzed to characterize the computer system.

Abstract: Some embodiments of the present invention provide a system that controls temperature variations in a computer system. During operation, a telemetry variable of the computer system is monitored. Next, a future temperature of the computer system is predicted based on the telemetry variable. A signal is then generated in response to the future temperature. Then, the signal is sent to a cooling device in the computer system to control temperature variations of the computer system.

Abstract: A method of simulating a design described in HDL is provided. In this method, modules of the design can be partitioned into first modules for simulation by a serial simulation engine and second modules for simulation by a concurrent simulation engine. The first and second modules can be prioritized for simulation based on classes of events consistent with an execution model of the HDL. Simulations of the serial and concurrent simulation engines can be synchronized for each class of events. Synchronizing can include transferring updated interface variable values, which are shared by the second modules and at least a subset of the first modules, between the serial simulation engine and the concurrent simulation engine. This transferring can include translating representations of the updated interface variable values.

Abstract: Embodiments of a system that adjusts a checkpointing frequency in a distributed computing system that executes multiple jobs are described. During operation, the system receives signals associated with the operation of the computing nodes. Then, the system determines risk metrics for the computing nodes using a pattern-recognition technique to identify anomalous signals in the received signals. Next, the system adjusts a checkpointing frequency of a given checkpoint for a given computing node based on a comparison of a risk metric associated with the given computing node and a threshold, thereby implementing holistic fault tolerance, in which prediction and prevention of potential faults occurs across the distributed computing system.

Abstract: Simulating hardware includes generating a data flow representation of the hardware, based on a hardware description language (HDL) description. The data flow representation including compatibility information that preserves behavioral and synthesizable characteristics of the HDL description. Simulating hardware further includes generating code from the data flow representation, and executing the code concurrently.

Abstract: A system that mitigates quantization effects in quantized telemetry signals. During operation, the system monitors a set of quantized telemetry signals. For a given quantized telemetry signal in the set of quantized telemetry signals, the system uses a set of models to generate a set of estimates for the given quantized telemetry signal from the other monitored quantized telemetry signals, wherein each model in the set of models was initialized using a different randomly selected subset of a training dataset. The system then averages the set of estimates to produce an estimated signal for the given quantized telemetry signal.

Abstract: A computer system to predict a value of a signal from a sensor schedule loads across a set of processor cores is described. During operation, the computer system generates N models to predict the value of the signal based on a set of quantized telemetry signals, where a given model produces a value of the signal using a subset of the set of quantized telemetry signals, and where the subset is selected from the set of quantized telemetry signals based on an objective criterion. Next, the computer system predicts the value of the signal by aggregating the values produced by the N models.

Abstract: A computer system that schedules loads across a set of processor cores is described. During operation, the computer system receives thermal measurements from sensors associated with the set of processor cores, and removes noise from the thermal measurements. Then, the computer system analyzes thermal properties of the set of processor cores based on the thermal measurements. Next, the computer system receives a process to be executed, and schedules the process to be executed by at least one of the processor cores based on the analysis. This scheduling is performed in a manner that reduces spatial and temporal thermal variations in the integrated circuit.

Abstract: One embodiment of the present invention provides a system that efficiently conducts vibrational characterizations for a computer system having variable component configurations. During operation, the system receives a given component configuration associated with the computer system. Next, the system looks up the given component configuration in a resonant spectra library, which contains structural resonant frequencies for a number of possible component configurations for the computer system. If the given component configuration is found in the resonant spectra library, the system retrieves a set of structural resonant frequencies associated with the given component configuration. The system subsequently controls one or more vibration sources within the computer system to avoid the set of structural resonant frequencies.

Abstract: One embodiment of the present invention provides a system for predicting a remaining useful life (RUL) for a computer system. The system starts by collecting values for at least one telemetry metric from the computer system while the computer system is operating. The system then uses the collected values to generate a RUL prediction for the computer system or a component within the computer system.

Abstract: Some embodiments of the present invention provide a system that characterizes a computer system using a pattern-recognition model. First, values for an environmental parameter are monitored from a set of sensors associated with the computer system. Then, a baseline for the environmental parameter is calculated based on the monitored values from a subset of the set of sensors. Next, the baseline is subtracted from the monitored values from sensors in the set of sensors to produce compensated values. Then, the compensated values are used as inputs to the pattern-recognition model, which produces estimates for the compensated values based on correlations between the compensated values learned during a training phase. Next, residuals are calculated by subtracting the estimates for the compensated values from the compensated values. Then, the residuals are analyzed to characterize the computer system.

Abstract: Embodiments of a system that adjusts a checkpointing frequency in a distributed computing system that executes multiple jobs are described. During operation, the system receives signals associated with the operation of the computing nodes. Then, the system determines risk metrics for the computing nodes using a pattern-recognition technique to identify anomalous signals in the received signals. Next, the system adjusts a checkpointing frequency of a given checkpoint for a given computing node based on a comparison of a risk metric associated with the given computing node and a threshold, thereby implementing holistic fault tolerance, in which prediction and prevention of potential faults occurs across the distributed computing system.

Abstract: A computer system that schedules loads across a set of processor cores is described. During operation, the computer system receives thermal measurements from sensors associated with the set of processor cores, and removes noise from the thermal measurements. Then, the computer system analyzes thermal properties of the set of processor cores based on the thermal measurements. Next, the computer system receives a process to be executed, and schedules the process to be executed by at least one of the processor cores based on the analysis. This scheduling is performed in a manner that reduces spatial and temporal thermal variations in the integrated circuit.

Abstract: A system that diagnoses a failure in a computer system is described. During operation, the system tests the computer system using a sequence of tests, where a given test includes a given load associated with a pre-determined failure mechanism for a given failure condition. During the given test, the system obtains results, which include telemetry signals that are monitored within the computer system. If the results indicate the given failure condition, the system ceases the testing and indicates that the computer system has the given failure condition. Otherwise, the system continues the sequence of tests until the sequence is completed, at which point, if no fault has been detected, the system indicates that a no-trouble-found (NTF) condition exists.