Ilya Grebnov

Computational tasks are mapped with computational locations in a distributed system such as a cloud computing environment. Mapping does not rely on workload estimates. Instead, tasks whose prerequisite tasks or other preconditions are determined to be mutually exclusive are co-located, while other tasks are mapped to different locations than one another. Locations are servers, processor cores, virtual machines, applications, or computational processes, for example. Mutual exclusivity may be… Show more

Computational tasks are mapped with computational locations in a distributed system such as a cloud computing environment. Mapping does not rely on workload estimates. Instead, tasks whose prerequisite tasks or other preconditions are determined to be mutually exclusive are co-located, while other tasks are mapped to different locations than one another. Locations are servers, processor cores, virtual machines, applications, or computational processes, for example. Mutual exclusivity may be determined by detecting that preconditions require different values of a shared variable in order to be satisfied, for example, or determining that preconditions correspond to different branches of a conditional programming statement. A satisfiability engine may also provide a satisfiability determination. Co-located tasks may also be batched, for improved execution performance. Co-location based on mutual exclusivity may result in fewer operations to save and restore task state, fewer cache misses, greater co-allocation of computational resources by the tasks, and easier debugging. Show less

Aspects of the present invention relate to guaranteeing delivery of replication messages in distributed storage systems. A request to perform an operation may be received. A first replication message corresponding to the request may be created, where the first replication message comprises instructions to replicate the result of the operation to one or more target storages. The replication message may be inserted in a message queue with a delayed visibility. The operation may be performed… Show more

Aspects of the present invention relate to guaranteeing delivery of replication messages in distributed storage systems. A request to perform an operation may be received. A first replication message corresponding to the request may be created, where the first replication message comprises instructions to replicate the result of the operation to one or more target storages. The replication message may be inserted in a message queue with a delayed visibility. The operation may be performed, where a result of the operation is persisted in a source storage in a data store. Upon determining that the result was successfully persisted in the source storage, a second replication message may be created, where the second replication message comprises instructions to replicate the result of the operation to the one or more target storages. The second replication message may be inserted in the message queue with an immediate visibility. Show less

Embodiments are directed to replicating data in distributed storage. A replication message may be retrieved from a message queue associated with a source table. The replication message may include a row identifier. One or more target storages within a same replication group as the source table may be identified. A row from each of the one or more target storages may be obtained corresponding to the row identifier. A winning row may be determined from the obtained rows based on a latest… Show more

Embodiments are directed to replicating data in distributed storage. A replication message may be retrieved from a message queue associated with a source table. The replication message may include a row identifier. One or more target storages within a same replication group as the source table may be identified. A row from each of the one or more target storages may be obtained corresponding to the row identifier. A winning row may be determined from the obtained rows based on a latest timestamp of the row. A replication operation may be created based on the winning row. The replication operation may be performed on the obtained rows from each of the target storages. Show less

Embodiments are directed to determining an optimal number of concurrently running cloud resource instances and to providing an interactive interface that shows projected operational metric measurements. In one scenario, a computer system accesses metric information which identifies operational metric measurements, and further accesses a second portion of metric information that identifies operational metric measurements for the cloud resource instances over a second period of time. The computer… Show more

Embodiments are directed to determining an optimal number of concurrently running cloud resource instances and to providing an interactive interface that shows projected operational metric measurements. In one scenario, a computer system accesses metric information which identifies operational metric measurements, and further accesses a second portion of metric information that identifies operational metric measurements for the cloud resource instances over a second period of time. The computer system then calculates projected operational metric measurements based on the identified operational metric measurements over the first period of time (e.g. for reactive tuning) and further based on the identified operational metric measurements over the second period of time (e.g. for predictive tuning) The computer system then determines, based on the projected operational metric measurements, a number of cloud resource instances that are to be concurrently running at a specified future point in time. Show less

Embodiments are directed to preventing flapping when auto-scaling cloud resources. The computer system calculates a scaling factor based on the target operational metric and the current measured value, where the scaling factor represents an amount of variance between the target operational metric and the current measured value. The computer system also calculates a delta value representing a modified quantity of cloud resources modified by the calculated scaling factor and determines whether a… Show more

Embodiments are directed to preventing flapping when auto-scaling cloud resources. The computer system calculates a scaling factor based on the target operational metric and the current measured value, where the scaling factor represents an amount of variance between the target operational metric and the current measured value. The computer system also calculates a delta value representing a modified quantity of cloud resources modified by the calculated scaling factor and determines whether a scaling action is to occur based on the calculated delta value. Show less