Browse Prior Art Database

System and Method for Power and Importance based Application Level Control in Data Center Disclosure Number: IPCOM000241796D
Publication Date: 2015-Jun-01
Document File: 2 page(s) / 51K

Publishing Venue

The Prior Art Database


Due to the varying power prices, it is important to alter the load of the data centers according the power price at that point in time. The normal approaches to reduce the power consumption in datacenter involve shutting down Virtual Machines and Physical Machines. This may result in loss of a complete application and running of other applications deems more important. Looking at applications from Virtual Machine/Physical machine level ignores the fact that every application composes of critical tasks and not so critical tasks. An approach that does not consider an application as a black box, and rather as a set of features/tasks will be able to provide much better experience to the customers.

This Cognitive system for smarter energy reduction in Data Centers uses Performance Testing Tool to do feature level energy profiling of applications (loaded in Data Centers) and correlating it with the importance factor of each feature (learnt from learning Models) to selectivity de-activate certain features of the application (by blocking its corresponding EHTTP-REQs), thereby maintaining the required energy state of the Data Center with minimal compromise on user experience.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 61% of the total text.

Page 01 of 2

System and Method for Power and Importance based Application Level Control in Data Center

Cognitive system for smarter energy reduction in Data Centers using the COMBINATION of the following methods:

Feature based energy profiling .

Each feature of an applications are first identified and their energy requirements are profiled using Performance testing solutions available. With this steps set of all the features and their energy requirements are calculated under realistic workloads. Calculated energy requirements are associated with respective features. Features to energy association is one of the key inputs to energy optimization. The step of feature energy profiling enables an application to be viewed as collection of energy consuming units as supposed to be a single energy consuming unit. Process of features energy profiling can be incremental, i.e. not every features have to energy profiled at the beginning but can be incremental.

Features prioritization.

A machine learning system is employed to observe application behavior over time for features being used. Output of this system is the predication of importance factor of each features at various time instance that are going to be used for energy optimization. Along with the system identified feature prioritization application owners can also provide the prioritization inputs. The final prioritization is combination of system identified and owners provided prioritization data.

EHTTP request identification