Abstract:
Technologies are generally provided for conveying tree and index needs of a database by collecting and forwarding a statistically representative set of data requests so that a new datacenter can use its existing auto-learning and adjustment techniques to create local indices before receiving its first real or live requests. In some examples, requests at an origin datacenter may be collected to provide a small population of queries that are representative of a larger population of queries. Filters anonymizers may be employed to adhere to data sharing rules and restrictions while still allowing a new datacenter to build indices and test performance. Thus, the new datacenter may be trained asynchronously from another site and without actual data changing hands. Training and testing of destination database services may be completed before committing to customer migration.
Abstract:
Technologies are generally described herein for managing heat within a phase change memory (PCM) structure utilizing electrocaloric effect materials. Some example PCM structures may include an electrocaloric effect material layer thermally coupled to an array of PCM cells. The electrocaloric effect material layer may be segmented so that activation of each segment is coordinated with a subset of the PCM cells within the array. While excess heat emanates from a PCM cell during memory operations, a corresponding electrocaloric effect material segment may be activated to decrease the thermal resistance of the electrocaloric effect material, which transfers the excess heat away from the neighboring PCM cells.
Abstract:
Technologies related to migration in place are generally described. In some examples, software and/or data associated with a platform customer can be “migrated” to a new platform while allowing the platform customer's software and/or data to remain on same hardware. A data center and/or Infrastructure as a Service (IaaS) provider may be configured to support migration in place for example by providing hardware identifiers upon request, providing platform identifiers upon request, providing platform compatibility information upon request and/or approving platforms for migration in place operations, deploying platform modules to migration in place hardware, providing user session support during migration in place operations, and/or performing security operations.
Abstract:
Technologies are generally described for pausing virtual machines using API signaling. In some examples, active threads in a VM may be listed on a registered thread list. When the active threads in the VM perform external API calls, they may provide wait messages, which may then be compared with the registered thread list to determine if the VM should be paused. Subsequently, one or more replies to the external API calls may be received and used to determine if the VM should be reactivated.
Abstract:
Technologies are generally provided for progressive key value store switchover by evaluating a maturity of a migrated data store and allowing piecewise switching of substructure area query servicing from an origin data store to a destination data store. In some examples, abstractions of origin and destination tree structures may be compared to each other in order to generate an evaluation metric at substantially reduced performance evaluation load. The evaluation metric may target performance sampling while assuring a desired performance level with localized query servicing switchover. Piecewise data transfer may also be optionally enabled such that overall storage can be similar to the storage of a single data store copy while reducing an impact on existing data store services.
Abstract:
Technologies are generally described for time-correlating administrative events within virtual machines of a datacenter across many users and/or deployments. In some examples, the correlation of administrative events enables the detection of confluences of repeated unusual events that may indicate a mass hacking attack, thereby allowing attacks kicking network signatures to be detected. Detection of the attack may also allow the repair of affected systems and the prevention of further hacking, before the vulnerability has been analyzed or repaired.
Abstract:
Technologies are directed to prevention of malicious attacks through cartography of co-processors at a datacenter. According to some examples, configuration data to create a co-processor at a field programmable gate array (FPGA) may be received at a configuration controller. The configuration controller may determine unused arrangements for the co-processor and unused placements at the FPGA corresponding to the unused arrangements. The used arrangements and the unused placements, associated with a type of the co-processor, may be stored in a configuration matrix. One of the unused arrangements and one of the unused placements corresponding to the selected unused arrangement may then be selected by the configuration controller to create the co-processor.
Abstract:
Technologies are generally described to provide a leakage current variability based power management of a processor. According to some examples, instruction counters and aggregated power consumption of the processor may be used to process power measurements of the processor into linear equations. The linear equations may be processed to produce a set of leakage values for the processor. In an example scenario, computation data from a power controller and processor instruction counters (PICs) of a core of the processor may be used to determine the leakage current variability of the core. A table of linear combination samples may be generated from the computation data. A micro-architectural leakage map of the core may be generated from the linear combination samples within the table.
Abstract:
Technologies are directed to a communication of application microarchitecture attributes between datacenters. According to some examples, components of a distributed broker are executed at an origin datacenter and a destination datacenter to communicate the microarchitecture attributes of an application executed or being executed at the origin datacenter. The distributed broker may read an event counter register to capture an instruction counter of the application at the origin datacenter. The instruction counter may be added to a microarchitecture instruction census (MIC) at the origin datacenter. The MIC may be combined with an identification information of the application to produce a portable affinity record at the origin datacenter. In addition, the portable affinity record may be transmitted to the destination datacenter from the origin datacenter.