Abstract: In an example, a method of placing a virtual computing instance among a plurality of datacenters includes comparing constraints specified for the virtual computing instance against resource metrics obtained from the plurality of datacenters, and a budget specified for the virtual computing instance against admission prices for a plurality of tiers of the plurality of datacenters, to generate a candidate set of datacenters. The method further includes scoring the candidate set of datacenters. The method further includes deploying the virtual computing instance to a target datacenter selected from the candidate set of datacenters that satisfies a threshold score.

Abstract: A method for scheduling computing resources without container migration includes determining a resource availability for one or more hosts, a resource allocation for one or more virtual machines (VMs), and a resource usage for one or more containers. The method further includes calculating a target resource configuration for one or more VMs, wherein calculating a target resource configuration comprises determining an upper limit of resource demand on a VM from one or more containers allocated on the VM, based at least in part on the resource usage. The method also includes removing or adding resources to each of the one or more VMs for which a target resource configuration was calculated to achieve the target resource configuration for each VM. The method further includes allocating the one or more VMs on the one or more hosts based on the resource availability of the one or more hosts.

Abstract: A method for scheduling computing resources with container migration includes determining a resource availability for one or more hosts, a resource allocation for one or more virtual machines (VMs), and a resource usage for one or more containers. The method includes identifying the hosts on which VMs and containers can be consolidated based on resource availability. The method also includes calculating a target resource configuration for one or more VMs. The method further includes removing or adding resources to the VMs for which a target resource configuration was calculated to achieve the target resource configuration. The method further includes allocating the one or more VMs on the one or more hosts based on the resource availability of the one or more hosts, and allocating the one or more containers on the one or more VMs based on the resource configuration of each VM and the resource usage of each container.

Abstract: A system and method for allocating power resources among host computers in a cluster uses lower and upper bounds with respect to a power budget to be distributed to each of the hosts. Each host is allocated a portion of the cluster power capacity. Any excess amount of the capacity is then allocated to the hosts based at least partly on the lower bound (reserve capacity) and the upper bound (host power limit) of each of the clients.

Abstract: A method of managing virtual resources executing on a hardware platform that employs sensors to monitor the health of hardware resources of the hardware platform, includes filtering sensor data from the hardware platform and combining the sensor data with a fault model for the hardware platform to generate a health score, receiving an inventory that maps the virtual resources to the hardware resources of the hardware platform, receiving resource usage data describing use of the hardware resources of the hardware platform by the virtual resources, and generating resource utilization metrics from the resource usage data. The method includes receiving policy data specifying rules applicable to the inventory, determining a set of recommendations for changes to the inventory based on the health score, the resource usage data, and the policy data, and executing at least one recommendation to implement the changes to the inventory.

Abstract: An automatic scaling system and method for reducing state space in reinforced learning for automatic scaling of a multi-tier application uses a state decision tree that is updated with new states of the multi-tier application. When a new state of the multi-tier application is received, the new state is placed in an existing node of the state decision tree only if a first attribute of the new state is same as a first attribute of any state contained in the existing node and a second attribute of the new state is sufficiently similar to a second attribute of each existing state contained in the existing node based on a similarity measurement of the second attribute of each state contained in the existing node with the second attribute of the new state.

Abstract: A system and method for performing a resource allocation diagnosis on a distributed computer system includes obtaining a target resource allocation and a snapshot of the distributed computer system, where the snapshot includes configurations and resource usage information of at least some components of the distributed computer system, and generating a resource allocation recommendation based on the target resource allocation and the snapshot by iteratively traversing a resource hierarchy in the distributed computer system. The resource allocation recommendation specifies at least one resource configuration action or at least one capacity expansion action for the distributed computer system to meet the target resource allocation.

Abstract: A resource management system and method for automatically creating affinity-type rules for resource management in a distributed computer system uses association inference information for at least one resource to determine resource association between resources, which is used to automatically create an affinity-type rule for the resources. The affinity-type rule is considered when executing a resource management operation.

Abstract: A resource management system and method for performing resource capacity management in a cluster of host computers uses a snapshot of the cluster with one or more ghost host computers added to the cluster to execute a power management analysis. A ghost host computer is a fictitious construct based on a physical host computer. The results of the power management analysis may then be used as a cluster capacity recommendation to increase resource capacity of the cluster of host computers.

Abstract: A system and method for performing a hypothetical power management analysis on a distributed computer system uses chronologically consecutive snapshots of the distributed computer system. The snapshots are used to extract demands of clients running in the distributed computer system for a resource for different time intervals, which are then stitched together to produce a workload trace. The snapshots and the workload trace are used to construct modeling scenarios for the distributed computer system. The modeling scenarios are used to perform analyzes to simulate the operation of the distributed computer system during which the power management module is enabled to compute potential power savings.

Abstract: Application resource scheduler module is provided to achieve cooperative application workload scheduling for a consolidated virtual environment. The application resource scheduler aids an application workload scheduler that is part of a distributed computing application, such as Hadoop, to achieve a specified relative priority of the application workload virtual machines to other virtual machines in the virtual environment. The application resource scheduler assists in achieving cooperative workload scheduling by revising the amount of resources that the application workload scheduler sees as available and by setting resource controls for the virtual machines of the distributed computing application to influence the resources the virtual machines receive from the underlying consolidated virtual environment.

Abstract: A management system and method for remediating poor-performing clients running in a distributed computer system uses a machine learning technique to automatically detect one or more poor-performing clients among a plurality of clients running in the distributed computer based on at least performance data and resource usage data of the clients. An action is then initiated to mitigate the effects of the poor-performing clients.

Abstract: A system and method for performing an operational metric analysis for a virtual appliance uses application operational data from multiple instances of the virtual appliance. The application operational data is then used to generate an operational metric prediction for the virtual appliance.

Abstract: Exemplary methods, apparatuses, and systems include a first host system configuring storage of the first host to serve as a primary cache for a virtual machine running on the first host. A second host system configures storage of the second host to serve as a secondary cache and boots a placeholder virtual machine. The first host transmits, in response to write operations from the virtual machine directed to the primary cache, copies of the write operations to the second host to create mirrored copies on the secondary cache. The first host acknowledges each write operation from the virtual machine when the write operation is committed to both the primary cache and the secondary cache. When the virtual machine is restarted on the second host in response to a failure or migration event, the secondary cache is promoted to serve as a new primary cache for the virtual machine.

Abstract: A management server and method for performing resource management operations in a distributed computer system uses at least one sampling parameter to estimate demand of a client for a resource. The sampling parameter has a correlation with at least one target performance goal of an application that the client is running. The demand estimation can then be used to make at least one decision in a resource management operation.

Abstract: Exemplary methods, apparatuses, and systems include a first host system configuring storage of the first host to serve as a primary cache for a virtual machine running on the first host. A second host system configures storage of the second host to serve as a secondary cache and boots a placeholder virtual machine. The first host transmits, in response to write operations from the virtual machine directed to the primary cache, copies of the write operations to the second host to create mirrored copies on the secondary cache. The first host acknowledges each write operation from the virtual machine when the write operation is committed to both the primary cache and the secondary cache. When the virtual machine is restarted on the second host in response to a failure or migration event, the secondary cache is promoted to serve as a new primary cache for the virtual machine.

Abstract: A distributed computing application is described that provides a highly elastic and multi-tenant platform for Hadoop applications and other workloads running in a virtualized environment. Multiple instances of a distributed computing framework, such as Hadoop, may be executed concurrently. A centralized manager detects when contention for computing resources, such as memory and CPU, causes tasks to run slower on VMs executing on a given host, and scales up or scales down a cluster based on the detected resource contention.

Abstract: One embodiment of the system disclosed herein facilitates reduction of latency associated with accessing content of a memory page that has been swapped out by a guest operating system in a virtualized computer system. During operation, a hypervisor detects an I/O write command issued by the guest operating system at a swap location within the guest operating system's swap file and records the swap location. The hypervisor then prefetches contents of a page stored at the swap location within the guest operating system's swap file into a prefetch cache in host machine memory. Subsequently, the hypervisor detects an I/O read command issued by the guest operating system at the swap location within the swap file. In response, the hypervisor provides contents of the page to the guest operating system from the prefetch cache, thereby avoiding accessing the guest operating system's swap file.

Abstract: A system and method for performing a hypothetical power management analysis on a distributed computer system uses chronologically consecutive snapshots of the distributed computer system. The snapshots are used to extract demands of clients running in the distributed computer system for a resource for different time intervals, which are then stitched together to produce a workload trace. The snapshots and the workload trace are used to construct modeling scenarios for the distributed computer system. The modeling scenarios are used to perform analyses to simulate the operation of the distributed computer system during which the power management module is enabled to compute potential power savings.

Abstract: A system and method for allocating power resources among host computers in a cluster uses lower and upper bounds with respect to a power budget to be distributed to each of the hosts. Each host is allocated a portion of the cluster power capacity. Any excess amount of the capacity is then allocated to the hosts based at least partly on the lower bound (reserve capacity) and the upper bound (host power limit) of each of the clients.