公开(公告)号：US20110302286A1

公开(公告)日：2011-12-08

申请号：US12792400

申请日：2010-06-02

Applicant: Luca Parolini ,  Niraj Tolia ,  Bruno Sinopoli ,  Bruce Krogh

Inventor： Luca Parolini ,  Niraj Tolia ,  Bruno Sinopoli ,  Bruce Krogh

IPC: G06F15/173

CPC classification number: G06F9/5094 ,  Y02D10/22

Abstract: A system for managing computing nodes and cooling devices cooling the computing nodes in an infrastructure includes one or more managers. The one or more managers receive a service request for a workload to be run on one or more of the computing nodes. Candidate workload placements for placing the workload on the one or more computing nodes are determined. A first cost associated with each candidate workload placements using a computational network model. A second cost associated with the cooling devices for each candidate workload placement is determined using a thermal network model. A cost function is minimized. The cost function determines an economic cost of operating the infrastructure while running the requested workload and is based upon the first cost, the second cost. A candidate workload placement is selected from the candidate workload placements based on the minimized cost function.

Abstract translation: 用于管理计算节点和冷却设备的系统用于冷却基础设施中的计算节点包括一个或多个管理器。 一个或多个管理器接收在一个或多个计算节点上运行的工作负载的服务请求。 确定将工作负载放置在一个或多个计算节点上的候选工作负载布局。 使用计算网络模型与每个候选工作量布局相关联的第一个成本。 使用热网络模型确定与用于每个候选工作负载布局的冷却装置相关联的第二成本。 成本函数最小化。 成本函数确定在运行所请求的工作量时运行基础设施的经济成本，并基于第一个成本，第二个成本。 基于最小化成本函数，从候选工作负载布局中选择候选工作负载布局。

公开(公告)号：US08539059B2

公开(公告)日：2013-09-17

申请号：US12792400

申请日：2010-06-02

Applicant: Luca Parolini ,  Niraj Tolia ,  Bruno Sinopoli ,  Bruce Krogh

Inventor： Luca Parolini ,  Niraj Tolia ,  Bruno Sinopoli ,  Bruce Krogh

IPC: G06F15/173

CPC classification number: G06F9/5094 ,  Y02D10/22

Abstract: A system for managing computing nodes and cooling devices cooling the computing nodes in an infrastructure includes one or more managers. The one or more managers receive a service request for a workload to be run on one or more of the computing nodes. Candidate workload placements for placing the workload on the one or more computing nodes are determined. A first cost associated with each candidate workload placements using a computational network model. A second cost associated with the cooling devices for each candidate workload placement is determined using a thermal network model. A cost function is minimized. The cost function determines an economic cost of operating the infrastructure while running the requested workload and is based upon the first cost, the second cost. A candidate workload placement is selected from the candidate workload placements based on the minimized cost function.

Abstract translation: 用于管理计算节点和冷却设备的系统用于冷却基础设施中的计算节点包括一个或多个管理器。 一个或多个管理器接收在一个或多个计算节点上运行的工作负载的服务请求。 确定将工作负载放置在一个或多个计算节点上的候选工作负载布局。 使用计算网络模型与每个候选工作量布局相关联的第一个成本。 使用热网络模型确定与用于每个候选工作负载布局的冷却装置相关联的第二成本。 成本函数最小化。 成本函数确定在运行所请求的工作量时运行基础设施的经济成本，并基于第一个成本，第二个成本。 基于最小化成本函数，从候选工作负载布局中选择候选工作负载布局。