Browse Prior Art Database

Method and system for optimal workload placement in a virtualized IT environment based on neural networks #gtscomp15 Disclosure Number: IPCOM000246832D
Publication Date: 2016-Jul-05
Document File: 5 page(s) / 65K

Publishing Venue

The Prior Art Database


The increasing complexity of IT makes difficult to dynamically assign the best IT resources to support workloads, which continuously change their characteristics. Consequently, the optimization process usually requires manual and time-consuming activities, whereas automatic processes are generally available only for simple environments. This invention leverages the artificial neural network model to provide an automatic way to calculate the best mapping between available IT resources and a given list of workloads.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 39% of the total text.

Page 01 of 5

Method and system for optimal workload placement in a virtualized IT environment based on neural networks #gtscomp15

    In the context of a virtual environment, a crucial need is the continuous optimization of resources allocation, in order to provide the agreed service level for each workload, keeping the investment sustainable. Achieving the above objective is complex, especially when:

    There is a high density of workloads running into the virtual environment
The workloads frequently change their requirements in terms of throughput, usage of CPU, usage of memory, etc
The above scenario and needs are particularly common for cloud environments, especially for Platform as a Service (PaaS) and Software as a Service (SaaS) models.

    Infrastructure as a Service (IaaS) model is the easiest use case, because the cloud users buy a specific amount of virtual resources for each service they need to use (compute, storage, network, etc …). Consequently, the cloud provider has to optimize the management operations for the underneath physical resources of the virtualization environment. The virtualization technologies allow the cloud provider to allocate virtual resources through a cluster of physical resources, considering their actual usage and allowing a certain level of resources over-commitment. This approach allows to optimize costs, while the monitoring of the real performance of services provides the way to control the achievement of committed service level objectives.

    In case of a PaaS or SaaS, the cloud users buy a more complex capability, for example to develop and run applications. Customer is interested in specifying functional and non-functional requirements including performance (for example in terms of transaction response time), capacity (for example in terms of number of transaction or concurrent users), availability and resiliency. In this context, only the cloud provider knows the infrastructure resources, composed by a topology of virtual machines and/or containers running the application workloads. Hence, the cloud provider has more control and more responsibilities over the optimization of cloud resources usage, in order to minimize costs while keeping the required service level objectives of every cloud service.

    Due to the increasing IT complexity and dynamic nature of the workloads (in terms of their requirements), it is more and more complex the continuous optimization of system resource allocation to satisfy the expectation in terms of declared service level objectives. Indeed the optimization processes usually requires time-consuming manual activities, whereas optional automatic processes are generally available only for simple environments, without focusing on protecting the return of investment (ROI) of acquired IT resources, as well as without the ability to guarantee the declared service level objectives of the running workloads.

    In order to address the above goals, this invention provides a method and system to calc...