Browse Prior Art Database

A Cognitive Approach to Moving Applications in the Cloud

IP.com Disclosure Number: IPCOM000246210D
Publication Date: 2016-May-17
Document File: 2 page(s) / 75K

Publishing Venue

The IP.com Prior Art Database

Abstract

A method for applying learning algorithms to multi-site geographically dispersed (probably worldwide) cloud environments. The learning approach draws from information sources internal to the cloud as well as external factors not generally considered by todays cloud optimisation routines. The result is the new ability to decide whether to move an application to a different cloud data center, potentially in a different geography, based upon all the factors that are important to the customer.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 47% of the total text.

Page 01 of 2

A Cognitive Approach to Moving Applications in the Cloud

The major players in the global cloud computing market (such as IBM, Amazon, HP, Oracle, etc) have multiple cloud data centres in known locations around the world. When deploying applications to the cloud a choice must be made, either manually or automatically, for the physical location of where to run that application. The customer may have some influence over this choice, for example if they need to service requests from a particular geography or if for some reason their data must stay within a certain geography.

    Described is a learning method that takes into account a large set of parameters in order to recommend a location (from a given known set of cloud data centres) where an application should run or recommend the run specification for the application in a given data centre. This recommendation can either be automated such that apps are automatically deployed in the location or configuration recommended or the recommendation could be provided to a user (either a client user or cloud provider admin) who could manually pick the location or configuration.

    Proposed is a system that learns from a set of key operational parameters for running an application in the cloud in order to recommend a data centre in which to run or scale an application that would optimise one or more of these operational parameters. These key operational parameters could, for example, take into account the geography of the app users, the time zone of the users, wholesale electrical costs, weather, exchange rates, etc. Such a system has the ability to continually assess a cloud application and the infrastructure on which it runs with the view of moving or scaling the application, either automatically or manually after a prompt, between data centres in order maximise or minimise some objective on behalf of the cloud provider or application owner.

The system proposed has two modes of operation:


A learning mode - used for continually or periodically updating a model based upon


1.

key operational indicators


A query mode - used to continually or periodically determine if a given application


2.

should be moved to optimise the preferred operational indicators

1. Learning Mode
Key operational indicators are defined in advance of starting learning mode. However, they can also be changed dynamically while the system is learning. Each key operational indicator must be linked with a method of obtaining information to represent it. Example operational indicators might be but are not limited to:
wholesale electric cost at the location of each data centre: this information would be available locally to the cloud provider
expected application load at a given time of day
the current time of day
expected application load for certain events
the date, time and duration of certain events
the number of users of the application
the geography of the users of the application
a measure of the impact to users for each data centre (us...