Abstract: One or more embodiments of the present invention provide a technique for effectively managing virtualized computing systems with an unlimited number of hardware resources. Host systems included in a virtualized computer system are organized into a scalable, peer-to-peer (P2P) network in which host systems arrange themselves into a network overlay to communicate with one another. The network overlay enables the host systems to perform a variety of operations, which include dividing computing resources of the host systems among a plurality of virtual machines (VMs), load balancing VMs across the host systems, and performing an initial placement of a VM in one of the host systems.

Abstract: One or more embodiments of the present invention provide a method for performing initial placement and load balancing of a data objects in a distributed system. The distributed system includes hardware resources, e.g., host systems and storage arrays, which are configured to execute and/or store data objects. A data object is initially placed into the distributed system by creating a virtual cluster of hardware resources that are compatible to execute and/or host the data object, and then selecting from the virtual cluster a hardware resource that is optimal for executing and/or hosting the data object. The data object is placed into the selected hardware resource, whereupon a load balancing operation is optionally performed across the virtual cluster. The virtual cluster is subsequently released, and the distributed system is returned to its original state with the data object included therein.

Abstract: A system and method for performing automatic remediation in a distributed computer system with multiple clusters of host computers uses the same placement selection algorithm for initial placements and for remediation placements of clients. The placement selection algorithm is executed to generate a placement solution when a remediation request in response to a remediation-requiring condition in the distributed computer system for at least one client running in one of the multiple clusters of host computers is detected and a remediation placement problem for the client is constructed.

Abstract: Network bandwidth is allocated to virtual machines (VMs) according to a node hierarchy that includes a root node, intermediate nodes, and leaf nodes, wherein each leaf node represents a queue of packet transmission requests from a VM and each intermediate node represents a grouping of leaf queues. As VMs generate requests to transmit packets over the network, the network bandwidth is allocated by queuing packets for transmission in the leaf nodes, and selecting a leaf node from which a packet is to be transmitted based on tracking data that represent how much network bandwidth has been allocated to the nodes. Upon selecting the leaf node, the tracking data of the selected leaf node and the tracking data of an intermediate node that is a parent node of the selected leaf node are updated, and a command to transmit the packet of the selected leaf node is issued.

Abstract: Embodiments perform centralized input/output (I/O) path selection for hosts accessing storage devices in distributed resource sharing environments. The path selection accommodates loads along the paths through the fabric and at the storage devices. Topology changes may also be identified and automatically initiated. Some embodiments contemplate the hosts executing a plurality of virtual machines (VMs) accessing logical unit numbers (LUNs) in a storage area network (SAN).

Abstract: Systems and methods are disclosed that allow for transparently recovering from an uncorrected multi-bit error of arbitrary length located at a memory address. Storing one or more parity pages, for a set of pages in system memory, such that a page in the set of pages may be reconstructed using one of the parity pages is disclosed. Storing an indication of one or more page'disk location such that the one or more pages may be reconstructed by refilling the page from disk is also disclosed.

Abstract: Distributed storage resources having multiple storage units are managed based on data collected from online monitoring of workloads on the storage units and performance characteristics of the storage units. The collected data is sampled at discrete time intervals over a time period of interest, such as a congested time period. Normalized load metrics are computed for each storage unit based on time-correlated sums of the workloads running on the storage unit over the time period of interest and the performance characteristic of the storage unit. Workloads that are migration candidates and storage units that are migration destinations are determined from a representative value of the computed normalized load metrics, which may be the 90th percentile value or a weighted sum of two or more different percentile values.

Abstract: Distributed storage resources are managed based on data collected from online monitoring of workloads on the storage resources and performance characteristics of the storage resources. Load metrics are calculated from the collected data and used to identify workloads that are migration candidates and storage units that are migration destinations, so that load balancing across the distributed storage resources can be achieved.

Abstract: In one embodiment, a latency value is determined for an input/output IO request in a host computer of a plurality of host computers based on an amount of time the IO request spent in the host computer's issue queue. The issue queue of the host computer is used to transmit IO requests to a storage system shared by the plurality of host computers. The method determines a host specific value assigned to the host computer based in proportion on a number of shares assigned to the host in a quality of service policy for IO requests. The size for the host computer's issue queue is determined based on the latency value and the host specific value to control a number of IO requests that are added to the host computer's issue queue where other hosts in the plurality of hosts independently determine respective sizes for respective issue queues.

Abstract: Network bandwidth is allocated to virtual machines (VMs) according to a node hierarchy that includes a root node, intermediate nodes, and leaf nodes, wherein each leaf node represents a queue of packet transmission requests from a VM and each intermediate node represents a grouping of leaf queues. As VMs generate requests to transmit packets over the network, the network bandwidth is allocated by queuing packets for transmission in the leaf nodes, and selecting a leaf node from which a packet is to be transmitted based on tracking data that represent how much network bandwidth has been allocated to the nodes. Upon selecting the leaf node, the tracking data of the selected leaf node and the tracking data of an intermediate node that is a parent node of the selected leaf node are updated, and a command to transmit the packet of the selected leaf node is issued.

Abstract: One or more embodiments of the present invention provide a technique for effectively managing virtualized computing systems with an unlimited number of hardware resources. Host systems included in a virtualized computer system are organized into a scalable, peer-to-peer (P2P) network in which host systems arrange themselves into a network overlay to communicate with one another. The network overlay enables the host systems to perform a variety of operations, which include dividing computing resources of the host systems among a plurality of virtual machines (VMs), load balancing VMs across the host systems, and performing an initial placement of a VM in one of the host systems.

Abstract: An anomaly in a shared input/output (IO) resource that is accessed by a plurality hosts or clients is detected when a host that is not bound by any QoS policy presents large workloads to a shared IO resource that is also accessed by hosts or clients that are governed by QoS policy. The anomaly detection triggers a response from the hosts or clients as a way to protect against the effect of the anomaly. The response is an increase in window sizes. The window sizes of the hosts or clients may be increased to the maximum window size or in proportion to their QoS shares.

Abstract: A system and method for allocating a resource among clients running on host computers using capacity distribution uses lower and upper bounds with respect to a capacity to be distributed to each of the clients. Each client is allocated a portion of the capacity that corresponds to the lower bound for that client. Any excess amount of the capacity is then allocated to the clients based at least partly on the lower bound and the upper bound of each of the clients.

Abstract: A shared input/output (IO) resource is managed in a decentralized manner. Each of multiple hosts having IO access to the shared resource, computes an average latency value that is normalized with respect to average IO request sizes, and stores the computed normalized latency value for later use. The normalized latency values thus computed and stored may be used for a variety of different applications, including enforcing a quality of service (QoS) policy that is applied to the hosts, detecting a condition known as an anomaly where a host that is not bound by a QoS policy accesses the shared resource at a rate that impacts the level of service received by the plurality of hosts that are bound by the QoS policy, and migrating workloads between storage arrays to achieve load balancing across the storage arrays.

Abstract: In one embodiment, a latency value is determined for an input/output IO request in a host computer of a plurality of host computers based on an amount of time the IO request spent in the host computer's issue queue. The issue queue of the host computer is used to transmit IO requests to a storage system shared by the plurality of host computers. The method determines a host specific value assigned to the host computer based in proportion on a number of shares assigned to the host in a quality of service policy for IO requests. The size for the host computer's issue queue is determined based on the latency value and the host specific value to control a number of IO requests that are added to the host computer's issue queue where other hosts in the plurality of hosts independently determine respective sizes for respective issue queues.

Abstract: In one embodiment, a method receives current latency values from a plurality of host computers where a current latency value is calculated by a respective host computer based on an amount of time spent in the respective host computer's issue queue by an IO request most recently removed from the issue queue of the respective host computer. The issue queue of the respective host computer is used to transmit IO requests from the respective host computer to a storage system. The method then calculates a combined average latency value based on the current latency values and sends the combined average latency value to the plurality of host computers. Each respective host computer adjusts a size of the respective host computer's issue queue based on the combined average latency value, and the size controls a number of IO requests that are added to the respective host computer's issue queue.

Abstract: A shared input/output (IO) resource is managed in a decentralized manner. Each of multiple hosts having IO access to the shared resource, computes an average latency value that is normalized with respect to average IO request sizes, and stores the computed normalized latency value for later use. The normalized latency values thus computed and stored may be used for a variety of different applications, including enforcing a quality of service (QoS) policy that is applied to the hosts, detecting a condition known as an anomaly where a host that is not bound by a QoS policy accesses the shared resource at a rate that impacts the level of service received by the plurality of hosts that are bound by the QoS policy, and migration of workloads between storage arrays to achieve load balancing across the storage arrays.

Abstract: Systems and methods are disclosed that allow for transparently recovering from an uncorrected multi-bit error of arbitrary length located at a memory address. Storing one or more parity pages, for a set of pages in system memory, such that a page in the set of pages may be reconstructed using one of the parity pages is disclosed. Storing an indication of one or more page's disk location such that the one or more pages may be reconstructed by refilling the page from disk is also disclosed.

Abstract: One or more embodiments of the present invention provide a technique for effectively managing virtualized computing systems with an unlimited number of hardware resources. Host systems included in a virtualized computer system are organized into a scalable, peer-to-peer (P2P) network in which host systems arrange themselves into a network overlay to communicate with one another. The network overlay enables the host systems to perform a variety of operations, which include dividing computing resources of the host systems among a plurality of virtual machines (VMs), load balancing VMs across the host systems, and performing an initial placement of a VM in one of the host systems.

Abstract: A system and method for placing a client in a computer network system uses continuously variable weights to resource utilization metrics for each candidate device, e.g., a host computer. The weighted resource utilization metrics are used to compute selection scores for various candidate devices to select a target candidate device for placement of the client.