摘要:
Resource allocations in stream data processing systems are performed to allocate the resources in a cost-effective manner by formulating the resource allocation problem as a linear program. For a problem with a single output, a backward algorithm or method is used that produces an optimal solution in linear time. For a problem with multiple outputs and the network has a tree structure, a backward shrink algorithm or method is used to provide an optimal solution in linear time. These algorithms are fully distributed, they require only local exchange of parameters between neighboring nodes, and are adaptive to the dynamic changes in network conditions and flow rate fluctuations.
摘要:
Resource allocations in stream data processing systems are performed to allocate the resources in a cost-effective manner by formulating the resource allocation problem as a linear program. For a problem with a single output, a backward algorithm or method is used that produces an optimal solution in linear time. For a problem with multiple outputs and the network has a tree structure, a backward shrink algorithm or method is used to provide an optimal solution in linear time. These algorithms are fully distributed, they require only local exchange of parameters between neighboring nodes, and are adaptive to the dynamic changes in network conditions and flow rate fluctuations.
摘要:
A system and method for performance modeling for an information technology (IT) system having a server(s) for performing a number of types of transactions includes receiving data for system topology and transaction flows and receiving performance measurement data for the IT system. The measurement data is clustered into multiple regimes based on similarities. Service demand and network delay parameters may be inferred based on clustered data.
摘要:
A method for constructing an overlay multicast tree to deliver data from a source to an identified group of nodes is provided in which a plurality of nodes are identified and mapped into multidimensional Euclidean space. A geometric region is constructing having a size that is the minimum size necessary to contain the source and all the nodes. Once constructed, a tree is created beginning at the source and including all of the nodes within the geometric region.
摘要:
Methods and apparatus operating in a stream processing network perform load shedding and dynamic resource allocation so as to meet a pre-determined utility criterion. Load shedding is envisioned as an admission control problem encompassing source nodes admitting workflows into the stream processing network. A primal-dual approach is used to decompose the admission control and resource allocation problems. The admission control operates as a push-and-pull process with sources pushing workflows into the stream processing network and sinks pulling processed workflows from the network. A virtual queue is maintained at each node to account for both queue backlogs and credits from sinks. Nodes of the stream processing network maintain shadow prices for each of the workflows and share congestion information with neighbor nodes. At each node, resources are devoted to the workflow with the maximum product of downstream pressure and processing rate, where the downstream pressure is defined as the backlog difference between neighbor nodes. The primal-dual controller iteratively adjusts the admission rates and resource allocation using local congestion feedback. The iterative controlling procedure further uses an interior-point method to improve the speed of convergence towards optimal admission and allocation decisions.
摘要:
Techniques are provided for determining a capacity allocation in a multi-tiered computing system. In one aspect of the invention, a technique for designing capacity allocation for a multi-tiered computing system, each tier of the computing system having one or more computing devices (e.g., servers), comprises the following steps/operations. Input parameters relating to capacity allocation are obtained. A capacity allocation is computed based on at least a portion of the input parameters, the capacity allocation being computable such that one or more end-to-end performance measures are substantially satisfied for multiple service classes. The capacity allocation computation may minimize a cost function. The capacity allocation computation may also satisfy one or more mean delay guarantees for the multiple service classes, one or more tail distribution guarantees for the multiple service classes, or both one or more mean delay guarantees and one or more tail distribution guarantees for the multiple service classes.
摘要:
A method for constructing an overlay multicast tree to deliver data from a source to an identified group of nodes is provided in which a plurality of nodes are identified and mapped into multidimensional Euclidean space. A geometric region is constructing having a size that is the minimum size necessary to contain the source and all the nodes. Once constructed, a tree is created beginning at the source and including all of the nodes within the geometric region.
摘要:
A system and method for capacity sizing in a computer device or system includes determining one or more classes of operations based on at least one of historical computational usage and predicted usage for a system. Based on the one or more classes of operations, at least one capacity target is set based on the computational usage for each class such that computational capacity is maintained at a set level over a given time period and the set level satisfies at least one usage criterion over the given time period.
摘要:
Disclosed are a method of and a system for synchronizing clocks in a coordinated network of computers including a multitude of processing nodes, each of the nodes having a clock and one or more neighbor nodes. The method comprises the steps of electing one of the nodes as a correct leader node; and each of the non-leader nodes adjusting its clock rate, based on messages exchanged with neighbor nodes, to remain synchronized with the clock of said correct leader node. In a preferred embodiment, the adjusting step includes the step of each of the non-correct leader nodes using a weight assignment mechanism that gives neighbor nodes that are closer to the leader node more effect on the clock adjustment than those nodes that are further away from the correct leader node.
摘要:
The present invention is a method and an apparatus for predicting future travel times over a transportation network. In one embodiment, a method for predicting future travel times over a transportation network includes receiving a data point indicating a real-time volume of traffic on the link at a given time and updating a template representative of an observed traffic pattern on the link in accordance with the received data point. A future travel time over the link can then be estimated in accordance with the updated template. Thus, the template is able to adapt to dynamically changing traffic patterns, taking these changing traffic patterns into account when making predictions of future traffic patterns.