发明名称 |
Workload adaptive cloud computing resource allocation |
摘要 |
A workload associated with a task is assessed with respect to each of a plurality of computing paradigms offered by a cloud computing environment. Adaptive learning is employed by maintaining a table of Q-values corresponding to the computing paradigms and the workload is distributed according to a ratio of Q-values. The Q-values may be adjusted responsive to a performance metric and/or a value, reward, and/or decay function. The workload is then assigned to available computing paradigms to be performed with improved utilization of resources. |
申请公布号 |
US8793381(B2) |
申请公布日期 |
2014.07.29 |
申请号 |
US201213533164 |
申请日期 |
2012.06.26 |
申请人 |
International Business Machines Corporation |
发明人 |
Baughman Aaron K.;Boyer Linda M.;Codella Christopher F.;Darden Richard L.;Dubyak William G.;Greenland Arnold |
分类号 |
G06F15/173 |
主分类号 |
G06F15/173 |
代理机构 |
Hoffman Warnick LLC |
代理人 |
Lashmit Douglas A.;Hoffman Warnick LLC |
主权项 |
1. A cloud computing system having a computing resource including at least one computing device and providing a first service offering at least a first computing paradigm and a second computing paradigm, and a workload policy manager configured to identify a task to be assessed and to assign a workload associated with the task to at least one of the first computing paradigm or the second computing paradigm according to a resource allocation control method that configures the workload policy manager to:
initialize a table of Q-values for the task to be assessed and including a respective Q-value for each respective computing paradigm, each Q-value being set to a respective initial value; select one of the first computing paradigm or the second computing paradigm as a current computing paradigm for assessment of the task; determine at least one performance metric of the task to be assessed for the current computing paradigm; assess the at least one performance metric, wherein assessing the at least one performance metric includes: defining a target vector including a respective predefined threshold value for each of the at least one performance metric; defining a system performance vector for each computing paradigm, each system performance vector including a respective value of at least one respective performance metric; and comparing each system performance vector to the target vector; determine, responsive to the assessment of the performance metric, a respective change to be applied to a respective Q-value associated with each computing paradigm; apply the respective changes to the respective Q-values, including: increasing a respective Q-value associated with each computing paradigm responsive to the respective system performance vector being equal to or exceeding the target vector; and decreasing a respective Q-value associated with each computing paradigm responsive to the system performance vector being less than the target vector; and reassign a workload associated with the task among available computing resources based on a ratio between the Q-values. |
地址 |
Armonk NY US |