Browse Prior Art Database

System and Method to Predict and Dynamically Adjust the Allocation of Resources for Cloud

IP.com Disclosure Number: IPCOM000225612D
Publication Date: 2013-Feb-21
Document File: 8 page(s) / 399K

Publishing Venue

The IP.com Prior Art Database

Abstract

The emergence of cloud computing technology has produced a tremendous and profound impact on the deployment and operation of the application. When the workload of the application becomes high and the existing nodes will be overloaded, the cloud platform can create new nodes and deploy same environments with same pattern to split the flow.However, there may be some problems with the workload self adaptive method to make the resource utilization and deployment low and not effective. To attack the problems, in this disclosure, we’ll present a system and method for predictability and dynamically adjust the allocation of resources for cloud to promote resource utilization and avoid missing catch overload workload. Here are the details of the system and method.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 01 of 8

System and Method to Predict and Dynamically Adjust the Allocation of Resources for Cloud

The emergence of cloud computing technology has produced a tremendous and profound impact on the deployment and operation of the application. Different applications are deployed and running in the same cloud platform which can provide basic and necessary infrastructures and services to reduce maintenance difficulties of the application. One application may have several different nodes in which the app self and the supported middleware and OS are installed. The access requests from different clients will be distributed to the corresponding application by certain router system (Fig1). When the workload of the application becomes high and the existing nodes will be overloaded, the platform can create new nodes and deploy same environments with same pattern to split the flow.

Fig1A sample structure of application on cloud

Problems Solved:

However, there may be some problems with the workload self adaptive method. See the Fig2.

1



Page 02 of 8

Notes:
T1: The start time to prepare new node to handle the overload workload T2: The peak of the workload
T3: The workload return to the normal level
T4: The start time of starting to create and deploy
T5: The end time of creating and deploying

Fig2Deploy problem

Problem1: Just only one application and middleware will be deployed to the new image on the new created node (1:1:1). Resource utilization seems a little low. Problem2: Usually creating a new image with ready application and related supported environments will take longer time (T5-T4) than the "workload necessary time" (T3-T1). That means it's even possible that the workload have been down to the normal level from the peak while the new image creating and deploying work may be not finished. The situation may result in the overload workload cannot be handled and users' requesting may be refused.

Problem3:When create and deploy a number of VMs it will be much harder due to concurrency and resource competition on workload demand.

To attack the above problems, in this disclosure, we'll present a system and method for predictability and dynamically adjust the allocation of resources for cloud to promote resource utilization and avoid missing catch overload workload. Here are the details of the system and method.

1. The workload and workload changes of the application nodes will be monitored by the workload monitor and a workload table of the nodes is built and the history data will be also saved to the data store for future use.

2. The system will check the compatibility of different applications by analyzing their pattern structure and related dependent properties or scripts to get the results which two or more applications compatible and which are not. These results will be used as pre-conditions in the later deploy policies.

2



Page 03 of 8

Fig3Application compatibilities check

3. A real-time predictor will analyze the above workload change history data w...