Amazon Technologies, Inc.

Vmware, Inc.

Agami Systems, Inc.

Nvidia Corporation

Zambeel, Inc.

William John Earl

Christos Karamanolis

Nachiketh Rao Potlapally

Rahul Gautam Patel

Eric Shannon Knauft

Pascal Renauld

John Merrill Phillips

Deepak Singh

Matthew Shawn Wilson

Kiran R Joshi

Mansi Shah

Nathan Burnett

David James Borland

Derek Del Miller

Uwe Dannowski

Kevin J Kranzusch

Terry Gene Hahn

Christopher W Johnson

Gregory A Smith

Satya Kiran Popuri

Dhanabal Ekambaram

Jeff Tofano

Embodiments of the disclosure provide techniques for measuring congestion and controlling quality of service to a shared resource. A module that interfaces with the shared resource monitors the usage of the shared resource by accessing clients. Upon detecting that the rate of usage of the shared resource has exceeded a maximum rate supported by the shared resource, the module determines and transmits a congestion metric to clients that are currently attempting to access the shared resource. Clients, in turn determine a delay period based on the congestion metric prior to attempting another access of the shared resource.

Distributed policy-based provisioning and enforcement for quality of service

Embodiments of the disclosure provide techniques for partitioning a resource object into multiple resource components of a cluster of host computer nodes in a distributed resources system. The distributed resources system translates high-level policy requirements into a resource configuration that the system accommodates. The system determines an allocation based on the policy requirements and identifies resource configurations that are available. Upon selecting a resource configuration, the distributed resources system assigns the allocation and associated values to the selected configuration and publishes the new configuration to other host computer nodes in the cluster.

Translating high level requirements policies to distributed storage configurations

Embodiments presented herein techniques for balancing a multidimensional set of resources of different types within a distributed resources system. Each host computer providing the resources publishes a status on current resource usage by guest clients. Upon identifying a local imbalance, the host computer determines a source workload to migrate to or from the resources container to minimize the variance in resource usage. Additionally, when placing a new resource workload, the host computer selects a resources container that minimizes the variance to further balance resource usage.

Load balancing of resources

A network-based virtual computing resource provider may offer virtual compute instances that implement rolling resource credits for scheduling virtual computing resources. Work requests for a virtual compute instance may be received at a virtualization manager. A resource credit balance may be determined for the virtual compute instance. The resource credit balance may accumulate resource credits in rolling fashion, carrying over unused credits from previous time periods. Resource credits may then be applied when generating scheduling instructions to provide to a physical resource to perform the work requests, such as a physical CPU in order to increase the utilization of the resource according to the number of credits applied. Applied resource credits may then be deducted from the credit balance.

Rolling resource credits for scheduling of virtual computer resources

Embodiments of the disclosure provide techniques managing a log-structured solid state drive (SSD) format in a distributed storage system. SSDs in the distributed storage system maintains a journal of logical changes to storage objects to persist prepared and committed changes in the latency path. The journal includes metadata entries that describe changes and reference data pages. Dense data structures (such as a logical block addressing table) index the metadata entries. To reduce the amount of overhead in I/O operations, the distributed storage system maintains the dense data structures in memory rather than on disk.

Log-structured storage device format

Translating high level requirements policies to distributed configurations

Embodiments of the disclosure provide techniques for updating a distributed transaction log on a previously offline resource object component using distributed transaction logs from active host computer nodes from separate RAID mirror configurations. Each component object maintains a journal (log) where distributed transactions are recorded. If a component object goes offline and subsequently returns (e.g., if the node hosting the component object reboots), the component object is marked as stale. To return the component object to an active state, a distributed resources module retrieves the journals from other resource component objects from other RAID configurations where the data is mirrored. The module filters corresponding data that is missing in the journal of the previously offline corresponding object and merges the filtered data to the journal.

Distributed transaction log

An embodiment of the invention sets forth a primary processing unit, a secondary processing unit coupled to the primary processing unit and accessible via a plurality of channels and a plurality of guest virtual machines executing on the primary processing unit. Each guest virtual machine includes a driver associated with the secondary processing unit, and a privileged virtual machine executing on the primary processing unit and configured to allocate a different set of channels of the plurality of channels to each of the drivers included in the guest virtual machines, where a first set of channels allocated to a first driver enables the first driver to access the secondary processing unit without conflicting with any of the other and with minimal performance overhead by directly accessing the secondary processing unit channels.

Paravirtualized virtual GPU

A multi-tenant trusted platform module (MTTPM) is attached to a communication bus of a virtualization host. The MTTPM includes a plurality of per-guest-virtual-machine (per-GVM) memory location sets. In response to an indication of a first trusted computing request (TCR) associated with a first GVM of a plurality of GVMs instantiated at the virtualization host, a first memory location of a first per-GVM memory location set is accessed to generate a first response indicative of a configuration of the first GVM. In response to an indication of a second TCR associated with a second GVM, a second memory location of a second-per-GVM memory location set is accessed to generate a second response, wherein the second response is indicative of a different configuration of the second GVM.

Multi-tenant trusted platform modules

Techniques are described for injecting noise in a timer value provided to an instruction requesting the timer value. A plurality of tasks may execute on a processor, wherein the processor may comprise one or more processing cores and each task may include a plurality of computer executable instructions. In accordance with one technique for injecting noise in the timer value, a request for a first timer value is received by one or more computer executable instructions belonging to a first task from the plurality of tasks, and in response, a second timer value is provided to the first task instead of the first timer value, wherein the second timer value is derived from the first timer value and a random number.

Noise injected virtual timer

Techniques are described for accumulating unused computing resources. The techniques may involve associating, with a task, a baseline amount of a computing resource for each time interval of a predetermined number of time intervals, and monitoring a consumption of the computing resource by the task in each time interval. Resource credits can be accumulated based on an unused amount of the computing resource during at least some of the time intervals. When a workload of the task consumes more than the baseline amount of the computing resource, the accumulated resource credits can be applied to allocate an additional amount of the computing resource to the task. A proportionate additional amount of a memory resource can also be allocated to the task.

Growing community of inventors

Burien, WA, United States of America

William John Earl