Browse Prior Art Database

A technique for enhancing the accuracy of predictive scaling decisions based on external business events

IP.com Disclosure Number: IPCOM000234676D
Publication Date: 2014-Jan-28
Document File: 3 page(s) / 79K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a technique for building up a set of data for a predictive scaling engine, to allow an application to be scaled to a required size in time to meet the predicted demands on the application for a business event. The data can be built up in a pre-production environment, to enable the predictive scaling engine to more accurately determine when to make scaling decisions once in production.

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

Page 01 of 3

A technique for enhancing the accuracy of predictive scaling decisions based on external business events

A predictive scaling engine can be used within a cloud infrastructure, to make predictions about the resource needs of an application to handle the workload generated by upcoming business events. To achieve this, the predictive scaling engine may monitor for upcoming events and use information sources such as a history of previous similar business events to predict the expected number of transactions per second on a particular business function of an application. This information can then be fed into a scaling algorithm, which uses past scaling history for the application to make a decision as to whether the application needs scaling to handle the anticipated demand.

The problem solved here, is that until the past scaling history is built up for the application, it is difficult to make a decision about the size of application deployment that would be needed to meet the required increase in demand on the application.

The application could be reactively scaled until enough information is available, but this introduces the risk of the application being overloaded until it has built up a reliable scaling history.

To build up the scaling history for the application, we need information for each business function of the application, as for example adding a new customer might be 1 DB2 call, where as querying a customers spending habits might cause 100s of DB2 calls and hence put more demand on a database.

Example table Business function of the
Application

Transactions Per Second Number of servers Number of database instances

AddCustomer 200

1

1

AddCustomer 400 1

1

AddCustomer 600 2

1

AddCustomer 800 3

2

AddCustomer 1000 4

2

AddCustomer 1150 Maximum TPS possible 4

2

QueryCustomerSpending 200 1

1

QueryCustomerSpending 400 2

2

QueryCustomerSpending 530 Maximum TPS possible 2

3

The intent of the proposed technique is to enable the customer to build up the required scaling history in a pre-production test environment. The scaling code is configured to scale the software re-actively once certain thresholds are reached. A program guides the user through a UI,...