Abstract: Method for manufacturing a heat sink for liquid cooling system of a compute blade of a supercomputer, comprising in particular a friction-stir welding step of a base plate at a machined central rib and at a machined rim.

Abstract: Cold plate for a supercomputer compute blade, said cold plate delimiting at least one opening configured to receive at least one heat sink configured to cool at least one electronic component, said cold plate comprising a cooling circuit, comprising channels within which is configured to circulate a “cold” heat transfer fluid to supply said at least one heat sink, and a discharge circuit, comprising channels within which a “hot” heat transfer fluid is configured to circulate after heating through the at least one heat sink, said cold plate consists of an assembly of several separate elements, wherein each pair of adjacent elements fluidly connected at a portion of the cooling circuit or of the discharge circuit comprises a sealing member at the interface of said connection.

Abstract: A method for scheduling computing tasks on a supercomputer including offline reinforcement learning (OFRL) of a scheduler on a database (LDB). The database includes at least one execution history (HIST) that includes the state (LHPCS) of a learning supercomputer at several moments (T, T?1); the actions (LACT) related to the scheduling of learning tasks on the learning supercomputer at those moments (T); and a reward (REW) related to each task. The method also includes the use of the scheduler trained on the computing tasks to be scheduled.

Abstract: A method for iteratively autotuning a high-performance computing system that depends on a set of parameters. Performance is first evaluated two or more times with the current values of the parameters. Afterward at least two evaluations, the median performance is evaluated. The median is then tested against a rule based on a filtering threshold. If the median does not the rule, the current values of the parameters are discarded, and the method is restarted with at least one other value generated by an optimization module; otherwise, a resampling method is performed based on the median and on a confidence interval that decreases with the number of steps of the optimization method.

Abstract: The invention relates to a device and a method for authenticating a user utilizing an internet access client (10) for accessing remote resources of a computer infrastructure, said access comprising a first authentication (130) of the internet access client (10) and a second authentication (140) of the user of the internet access client (10). The method includes sending (132), to a token security module (21), by the internet access client (10), a client certificate (220), said client certificate (220) being associated with items of identification information of the internet access client (10); and receiving (133), by the internet access client (10), an authentication token (210) generated by the token security module when the client certificate (220) sent has been verified by the token security module.

Abstract: An outlet heat sink for a liquid cooling system for an electronic board, including a cold inlet connector, a hot inlet connector and a hot outlet connector. An upper part of a cooling block, on the one hand, conveys the flow of heat transfer fluid entering through the cold inlet connector to the lower part of the cooling block so as to collect heat generated by the electronic board by being conveyed to the hot outlet connector, and on the other hand, conveys the flow of heated-up heat transfer fluid received on the hot inlet connector directly to the hot outlet connector.

Abstract: The invention relates to an edge-based log collecting device for collecting logs from several log sources located in a remote network, called edge-network, and sending the logs to a cloud-based system distant from the edge-network. The device is located in the edge-network, and the device includes several processing nodes for processing logs received from the log sources and sending them to said the system. The device is configured to elect, according to a predetermined election algorithm, one of the processing nodes as a master node configured for receiving the logs from all log sources of the local network, and sharing the logs with the other processing nodes. The invention further relates to a computer program and a device configured to carry out such a method.

Abstract: A computing blade for a liquid-cooled supercomputer, comprising as many network cards as there are processors, each network card being associated with a processor, each network card comprising a plug-in end configured to receive a connector, each network card being configured to allow data communication between said processor and another processor, each network card being mounted to the cold plate being connected to the motherboard via a connection system comprising a first connector, connected to the plug-in end of the network card, and a second connector, connected on the one hand to the first connector via a cable and on the other hand to the motherboard.

Abstract: An interlayer heat sink for a liquid cooling system for an electronic card, the interlayer heat sink including a cold inlet connector, a hot inlet connector, a hot outlet connector, and a cold outlet connector. An upper part of the cooling block divides the flow of heat transfer fluid entering through the cold inlet connector into a first flow, which is oriented towards the lower part of the cooling block so as to collect heat generated by the electronic card by being conveyed to the hot outlet connector, and a second flow, oriented directly to the cold outlet connector, to convey the flow of heated-up heat transfer fluid received on the hot inlet connector directly to the hot outlet connector where it fuses with the heated-up flow in the lower part of the cooling block.

Abstract: The invention relates to a method detecting one or more anomalies regarding logins of a user on an authentication system. The method includes at least one iteration of a monitoring phase that includes collecting a log of successful logins of the user on the authentication system, and calculating, for each monitored parameter, a probability density, in the log, of each predetermined value of the parameter. The method also includes calculating, for each parameter, a weight, as a function of the probability density of each predetermined value of the parameter, calculating an anomaly score for the log as a function of the weights and the probability densities; and comparing the anomaly score to a given threshold in order to determine an anomaly in the daily log for the user. The invention also relates to a computer program product and a device configured to carry out the method.

Abstract: Inlet heat sink for a liquid cooling system of an electronic board, comprising a “cold” inlet connector, a “hot” outlet connector, a “cold” outlet connector, the upper part of the cooling block being configured to divide the flow of incoming heat transfer fluid by the cold inlet connector into a first flow oriented to collect the heat generated by the electronic board by being routed to the hot outlet connector, and a second flow, oriented directly to the cold outlet connector.

Abstract: A method of representing a distributed computing system, the distributed computing system comprising a plurality of processing devices connected together according to a predefined topology. The method comprising receiving at least one piece of data from an activity log file relating to at least one processing device among the plurality of processing devices, receiving at least one metric relating to at least one processing device among the plurality of processing devices, receiving at least the predefined topology of the distributed computing system, constructing a graph representative of a distributed computing system operation, the graph comprising the data item extracted from the received log file, the received metric, and the received topology, and embedding at least one part of the graph to obtain at least one state vector representing the at least one part of the embedded graph.

Abstract: A method of broadcasting an information packet in a network including a sender of the packet and a plurality of recipients connected by a network equipment of a routing server, the routing server including a physical computing system and a software system, and the physical computing system operating alternatively between a kernel mode and a user mode. The software system controls the implementation of a step of receiving the packet, and, for each recipient, a step of duplicating the packet. The software system also controls the implementation of a step of routing the packet thanks to an IP address of the recipient, and a step of sending the packet to the recipient, wherein each step is executed by the physical computing system in kernel mode.

Abstract: The invention relates to a computer-implemented method for programming a checkpoint of at least one application executed on a computing infrastructure. The computing infrastructure includes a plurality of resources allocated to the execution of the application. The method includes determining a date of failure of the application as a function of a prediction of an anomaly of at least one resource of the plurality of resources, and programming the checkpoint as a function of the date of failure.

Abstract: The invention relates to a method for automatically analyzing a transaction log of a distributed computing system comprising a plurality of lines. The method includes, for each line, cutting the line into words, constructing a comparison vector by comparing the line with the other lines of the same size as the line, constructing a pattern from the comparison vector, and creating an event per pattern. The invention includes constructing a prediction model by training an artificial neural network on a group of training events, the prediction model being configured to predict the next event in the transaction log. The invention includes, for at least one event, using the prediction model to predict the event, from a group of prediction events, and generating from the prediction model, a causal graph of the event comprising a causal relation for each event of the group of prediction events responding to a relevance condition.

Abstract: A computing unit configured to be mounted in a HPC cabinet comprising an external body and an internal body. The internal body includes a bottom tray, bottom secondary electronic components, at least a bottom connector connected to the bottom secondary electronic components and a top tray being releasably coupled with the bottom tray. The top tray includes top secondary electronic components and at least a top connector connected to the top secondary electronic components, wherein the top connector is configured to cooperate with the bottom connector when the top tray is coupled with the bottom tray.

Abstract: Some embodiments are directed to a logging within a software application executed over an assembly of information processing devices. More particularly, some embodiments relate to a method allowing process logging in the case of a software application operating with several processes and/or threads.

Abstract: Examples include a method to position atoms. The method comprises considering a target Hamiltonian encoding a specific problem to resolve using an optical tweezer traps quantum computing system. The method also comprises considering a set of representative Hamiltonians function of a position configuration of atoms in the quantum computing system. The method further comprises determining a specific position configuration whereby a specific similarity measure between the target Hamiltonian and a specific Hamiltonian of the representative Hamiltonians function of the specific position configuration is improved compared to another similarity measure between the target Hamiltonian and at least one other representative Hamiltonian function of a position configuration differing from the specific position configuration.

Abstract: The invention relates to a method for training an artificial neural network (ANN) for object detection in images. The method includes a training phase of training the ANN with a training database (TDB) that includes training images, and measuring a performance of the trained ANN with a validation database that includes validation images. If the measured performance is not satisfactory, the method further includes an enhancing phase of detecting a mismatch between the validation images and a response of the trained ANN in response to the validation images, correlating the mismatch to a characteristic of the training images that impacts the training of the ANN, modifying the training database according to the characteristic, and training the ANN with the modified training database. The invention further relates to a non-transitory computer program and a device configured to carry out the method, and to an artificial neural network trained with the method.

Abstract: A method for monitoring a person of interest in video that may be generated by multiple cameras in different locations. A sequence of images along with data such as time and location, called a tracklet, may be generated for the person of interest. This reference tracklet may then be compared to other tracklets received at later times to determine if the person of interest appears in these other tracklets. Analyses may include locating persons in the video images, calculating person signatures, and comparing signatures to signatures of the person of interest. Analyses may also include filters for possible locations of the person of interest based on how quickly this person could move from one location to another, and filters based on color of the person of interest's clothing. Possible matches may be confirmed by an operator, while matches with high confidence may be automatically identified as the person of interest.