Microsoft Technology Licensing, LLC

Ishai Menache

Navendu Jain

Srikanth Kandula

Christian L Belady

Shane Kumar Mainali

James R Larus

Danny Allen Reed

Christian Herwarth Borgs

Jennifer Tour Chayes

Rushi Srinivas Surla

Sriram S Rao

Peter Bodik

Avrilia Floratou

Ashvin Agrawal

Joseph Naor

Christopher William Douglas

Virajith Jalaparti

Ilan Lobel

Joseph (Seffi) Naor

F Bruce Shepherd

Roy Schwartz

Hamid Nazerzadeh

Yang Lu

Mainak Ghosh

Ohad Shamir

Niv Buchbinder

Methods, systems, apparatuses, and computer program products are provided for prefetching data. A workload analyzer may identify job characteristics for a plurality of previously executed jobs in a workload executing on a cluster of one or more compute resources. For each job, identified job characteristics may include identification of an input dataset and an input bandwidth characteristic for the input dataset. A future workload predictor may identify future jobs expected to execute on the cluster based at least on the identified job characteristics. A cache assignment determiner may determine a cache assignment that identifies a prefetch dataset for at least one of the future jobs. A network bandwidth allocator may determine a network bandwidth assignment for the prefetch dataset. A plan instructor may instruct a compute resource of the cluster to load data to a cache local to the cluster according to the cache assignment and the network bandwidth assignment.

Cache and I/O management for analytics over disaggregated stores

In an embodiment, a partition cost of one or more of the plurality of partitions and a data block cost for one or more data blocks that may be subjected to a garbage collection operation are determined. The partition cost and the data block cost are combined into an overall reclaim cost by specifying both the partition cost and the data block cost in terms of a computing system latency. A byte constant multiplier that is configured to modify the overall reclaim cost to account for the amount of data objects that may be rewritten during the garbage collection operation may be applied. The one or more partitions and/or one or more data blocks that have the lowest overall reclaim cost while reclaiming an acceptable amount of data block space may be determined and be included in a garbage collection schedule.

Cost-based garbage collection scheduling in a distributed storage environment

Various technologies pertaining to scheduling network traffic in a network are described. A request to transfer data from a first computing device to a second computing device includes data that identifies a volume of the data to be transferred and a deadline, where the data is to be transferred prior to the deadline. A long-term schedule is computed based upon the request, wherein the long-term schedule defines flow of traffic through the network over a relatively long time horizon. A short-term schedule is computed based upon the long-term schedule, where devices in the network are configured based upon the short-term schedule.

Computing long-term schedules for data transfers over a wide area network

A system for managing allocation of resources based on service level agreements between application owners and cloud operators. Under some service level agreements, the cloud operator may have responsibility for managing allocation of resources to the software application and may manage the allocation such that the software application executes within an agreed performance level. Operating a cloud computing platform according to such a service level agreement may alleviate for the application owners the complexities of managing allocation of resources and may provide greater flexibility to cloud operators in managing their cloud computing platforms.

Resource management for cloud computing platforms

There is provided a method and system for process migration in a data center network. The method includes selecting processes to be migrated from a number of overloaded servers within a data center network based on an overload status of each overloaded server. Additionally, the method includes selecting, for each selected process, one of a number of underloaded servers to which to migrate the selected process based on an underload status of each underloaded server, and based on a parameter of a network component by which the selected process is to be migrated. The method also includes migrating each selected process to the selected underloaded server such that a migration finishes within a specified budget.

Process migration in data center networks

A data center system is described which includes multiple data centers powered by multiple power sources, including any combination of renewable power sources and on-grid utility power sources. The data center system also includes a management system for managing execution of computational tasks by moving data components associated with the computational tasks within the data center system, in lieu of, or in addition to, moving power itself. The movement of data components can involve performing pre-computation or delayed computation on data components within any data center, as well as moving data components between data centers. The management system also includes a price determination module for determining prices for performing the computational tasks based on different pricing models. The data center system also includes a 'stripped down' architecture to complement its use in the above-summarized data-centric environment.

Data center system that accommodates episodic computation

A method for adaptively allocating resources to a plurality of jobs. The method comprises selecting a first policy from a plurality of policies for a first job in the plurality of jobs by using a policy selection mechanism, allocating at least one resource to the first job in accordance with the first policy, and in response to completion of the first job, updating the policy selection mechanism to obtain an updated policy selection mechanism by using at least one processor. Updating the policy selection mechanism comprises evaluating the performance of the first policy with respect to the first job by calculating a value of a metric of utility for the first policy based on conditions associated with execution of the first job and updating the policy selection mechanism based on the calculated value and a delay of execution of the first job.

Allocation of computational resources with policy selection

This document describes techniques for dynamically placing computing jobs. These techniques enable reduced financial and/or energy costs to perform computing jobs at data centers.

Dynamically placing computing jobs

Growing community of inventors

Redmond, WA, United States of America

Ishai Menache