Abstract: In some examples, a computing system may gathering, from a machine learning unit associated with the computing system, data as a training data, label the data to identify the labeled data as the training data, which may be then recirculated for each of one or more analytics modules of the machine learning unit.

Abstract: Technologies and implementations for allocating datacenter resources are generally disclosed.

Abstract: Technologies for virtual machine placement within a data center are described herein. An example method may include determining a shared threat potential for a virtual machine based, at least in part, on a degree of co-location the virtual machine has with a current virtual machine operating on a physical machine, determining a workload threat potential for the virtual machine based, at least in part, on a level of advantage associated with placing the virtual machine on the physical machine, determining a threat potential for the virtual machine based, at least in part, on a combination of the shared threat potential and the workload threat potential, and placing the virtual machine on the physical machine based on the threat potential.

Abstract: Implementations and techniques for measuring and improving the quality of a user experience are generally disclosed.

Abstract: Technologies are generally described for suppressing discovery of mobile devices in a wireless ubiquitous computing (ubicomp) network, reducing the number of network transmissions by devices in the network and reducing network congestion and prolonging battery life. A mobile device may add supplementary information to introductory beacons transmitted by the device that indicate no response is necessary from other devices participating in the network except for a coordinator of the network. The mobile device may transmit a malformed introductory beacon that will be ignored by other devices in the network except for the coordinator. The coordinator of the wireless network may detect the mobile device moving through the network and transmit a suppression signal to other devices in the network causing the devices to ignore the introductory beacons transmitted by the mobile device.

Abstract: Technologies are generally described to provide a fail-safe licensing system. An example system may include a quorum detection module of an instance of a software program configured to query a licensing queue corresponding to the software program for messages to determine a number of active instances of the software program. Based on a comparison of the determined number of active instances and a maximum number of allowed instances of the software program, the quorum detection module may be configured to determine whether the instance of the software program is authorized. In response to a determination that the instance of the software program is authorized, the quorum detection module may generate a unique instance identification tag and post a message to the licensing queue, where the message includes an indication that the instance of the software program is active and an indication of the generated unique instance identification tag.

Abstract: Technologies are generally described for a system to process a collection of video recordings of a scene to extract and localize audio sources for the audio data. According to some examples, video recordings captured by mobile devices from different perspectives may be uploaded to a central database. Video segments capturing an overlapping portion of the scene at an overlapping time may be identified, and a relative location of each of the video capturing devices may be determined. Audio data for the video segments may be indexed with a sub-frame time reference and relative locations as a function of overlapping time. Using the indices that include the sub-frame time references and relative locations, audio sources for the audio data may be extracted and localized. The extracted audio sources may be transcribed and indexed to enable searching, and may be added back to each video recording as a separate audio channel.

Abstract: In one example embodiment, an instance of a virtual machine to evict may be identified. A determination may be made within a predetermined range of probabilities, that processing on the identified instance of the virtual machine will resume within a predetermined range of time. Thus, the identified instance of the virtual machine may be softly evicted by allocating processing resources away from the identified instance of the virtual machine, receiving an instruction to resume the processing on the identified instance of the virtual machine, and restoring the processing on the identified instance of the virtual machine.

Abstract: Techniques for displaying virtual objects on devices in differential situations are provided. Augmented reality authoring ensures that users have a consistent experience with virtual objects, including augmented reality images, by delivering the same versions of the virtual objects to devices that are close in terms, for instance, of at least distance and/or time. Devices that are not sufficiently close may receive different versions of the virtual object, thus ensuring that the users of devices that are sufficiently near each other do not experience different versions of the virtual object.

Abstract: Technologies related to virtual machine switching based on measured network delay are generally described. A network delay aware Virtual Machine (VM) may be configured to adapt a Virtual Machine Manager (VMM) to delay switching back to the network delay aware VM by a delay amount determined using a measured network delay. The measured network delay may comprise a delay between sending a network communication and receiving a network response. By delaying switching back to the network delay aware VM, additional processing resources are freed for other VMs managed by the VMM, thereby increasing efficiency of computing devices including network delay aware VMs, and correspondingly increasing efficiency of data centers including such computing devices.

Abstract: In one example embodiment, operations may include predicting winning bidders in an auction for computing resources; pre-placing machine images before the auction has been completed; booting-up at least a portion of the pre-placed machine images before the auction has been completed; and assigning a booted-up virtual machine to one of the predicted auction winners.

Abstract: Technologies are provided to automatically vary a structure of a netlist computation arranged to configure a field programmable gate array (FPGA). In an example scenario, an FPGA netlist may be received from a client to configure the FPGA. A perturbation generator may be activated in response to a detection of one or more security risk factors associated with the netlist. The netlist may be altered through schemes designed to repair one or more FPGAs. The repair schemes may be used to repair the FPGAs to work around failed cells and failed sub-cells. The perturbation generator may produce a false map of failed cells. The false map may be used to generate different timings and different intermediate values associated with the netlist to generate an alternate netlist. The alternate netlist may be used to configure the FPGA to prevent side channel attacks.

Abstract: Technologies are generally provided for a system to enhance security and prevent side channel attacks of targeted functions. Side channel attacks assume that the targeted functions operate at same speed each time, and observe timing data of the targeted functions to glean secure information. According to some examples, an enhanced security system may alter a processing speed of one or more subunits of a processor executing the targeted function(s) to transparently change an instantaneous performance of the processor in an unpredictable manner. The performance time of the targeted function(s) may thereby be randomized. A virtual machine manager (VMM) may identify a security risk for a targeted function, and trigger one or more subunits of the processor to operate at a reduced frequency. After completion of the targeted function, the subunits may be returned to a default performance speed.

Abstract: Technologies are generally described to monitor an ingress and egress of data to and from platform provided storage. In some examples, a data scrubbing certification module of a platform may be configured to determine an existence of a data retention and elimination policy of a service associated with an application executed at the platform, where the service may store application data within a data store of the platform. The data scrubbing certification module may activate a certification process for the application, and the data store may receive the application data inserted with one or more sentinels from the service such that the data scrubbing certification module may track the sentinels to verify an ingress and egress of the application data to and from the data store. Evidence that the service is compliant with the data retention and elimination policy may then be provided to the application based on the verification.

Abstract: Technologies related to observer filtered activity recommendations are generally described. In some examples, a computing device may collect activity and observer information for a device user, may identify limited observer activities, and may generate observer filtered activity recommendations for the device user, the observer filtered activity recommendations corresponding to the identified limited observer activities. Observer filtered activity recommendations may optionally be restricted in the presence of current observers outside a limited observer group. In some examples, the computing device may provide automated access to observer information and/or to observer filtered activity information to enable observer filtered activity recommendations by applications accessing the observer information and/or the observer filtered activity information.

Abstract: Technologies may be generally described to provide viewer optimized compression of a model. In some examples, a computing device may receive a request to compress a master model for a viewer. The computing device may determine shape primitives of the master model through use of a shape primitive identification technique such as a random sample consensus (RANSAC) technique. The identified or determined shape primitives may be subtracted from the master model to determine residues of the master model. A processed model may be generated from the residues of the master model and the shape primitives. Visible subsets, visible based on a view cone of the viewer, of the residues and the shape primitives may be selected from the processed model, from which a compressed model may be generated. The processed model may then be used to generate a second view without redetermining the shape primitives.

Abstract: Technologies are described for health monitoring using snapshot backups through test vectors. In some examples, health of an application deployed at a datacenter may be monitored and key metrics recorded in the metadata of progressive backup snapshots of an instance of the application such that warning metrics can be reviewed retrospectively upon failure of the instance and a snapshot can be automatically selected for restoration of the application instance based on lack of high incidence of suspect metric values. Moreover, an operating state associated with snapshot backups may be assessed as the snapshots are captured and selected ones with operating conditions desired as part of a test suite may be saved for use as test scenarios. In particular, state information from added or existing deployment monitoring may be used by a test logic process to evaluate whether each snapshot is needed for testing scenarios.

Abstract: In one example embodiment, live migration in a datacenter may include JIT compiling a process that is configured to be executed on both a source instruction set architecture and a destination instruction set architecture, mapping variables and address stacks of the process on both the source instruction set architecture and the destination instruction set architecture into a labeled form thereof, and mapping the labeled form of the variables and address stacks onto the destination instruction set architecture.

Abstract: Technologies are generally described for systems, devices and methods effective to execute virtual machines on a core. In some examples, a processor may execute a first virtual machine on the core and receive a first request for an identification of the core by the first virtual machine. The processor may generate a first identifier associated with the core and associate the first identifier with the first virtual machine. The processor may report the first identifier to the first virtual machine. The processor may further execute a second virtual machine on the core and receive a second request for an identification of the core by the second virtual machine. The processor may generate a second identifier associated with the core that may be different from the first identifier. The processor may associate the second identifier with the second virtual machine and report the second identifier to the second virtual machine.

Abstract: Technologies may be provided related to customization and delivery of atlases matched to hardware of a client device in real-time. In some examples, a system according to embodiments may perform an assessment of the target hardware, generate atlases matched to the target hardware, and modify application code to reference graphic assets in their new locations in the modified atlases. An increase in speed of application loading, an increase in speed of resource delivery, a reduction of memory consumption on the client device, and/or an increase in the client device frame rate may be achieved by matching the atlases to attributes of the video hardware.