Efficient Learning Machines: Theories, Concepts, and Applications for Engineers and System Designers by Rahul Khanna & Mariette Awad

Author:Rahul Khanna & Mariette Awad [Khanna, Rahul & Awad, Mariette]
Language: eng
Format: epub
Tags: Intelligence (AI) & Semantics, Machine Theory, Software Development & Engineering, Computers, General
ISBN: 9781430259893
Google: qPGQnAEACAAJ
Publisher: Apress
Published: 2015-04-30T20:42:45+00:00

Mutation: Antigen–antibody interaction, coupled with somatichypermutation, forms the basis of an AIS. Mutation introduces diversity in the population and facilitates effective response to antigens.

AIS uses an adaptive population of antibodies to facilitate intelligent behavior by synthesizing diverse subset solutions for a given problem domain. AIS has been applied in areas related to network security and anomaly detection.

Distributed Management in Datacenters

Datacenters are complex environments that deal with key challenges related to power delivery, energy consumption, heat management, security, storage performance, service assurance, and dynamic resource allocation. These challenges relate to providing effective coordination to improve the stability and efficiency of datacenters. The fluctuating demands and diverse workload characteristics of a large datacenter make complex the tasks of upholding workload performance, cooling efficiency, and energy targets (discussed in the following sections). In such large clusters of systems, multiple objectives compete to accomplish service-level goals by avoiding actuator overlapping and exhausting a complex combination of constraints, timing granularity, type of approach, and sequence of controls. However, the combinatorial solution space can be extremely large and may not converge to a global optimal in a bounded time. Therefore, a centralized datacenter management system may not scale well in constrained time and hence may not deliver an optimal management solution.

SI has emerged as a promising field that can be exploited to construct a distributed management methodology leading to scalable solutions without centralized control. The following sections present a control system that identifies suitable targets for workload placement, with these fundamental control elements:

Controlled process: The controlled process implements the feedback control loop, which constrains the temperature and power of compute clusters, such as server racks, for a given policy. An optimal process operates within policy constraints and provisions sufficient energy to operate a workload at highest performance efficiency and lowest cooling.

Fitness function: The fitness function estimates the most favorable placement of the workload, based on the existing knowledge base’s expected demand and availability of resources.

Knowledge base: The knowledge base acts as a finite database made up of survey data conducted by the sensor agents. This knowledge assists in identifying the most probable placement of the workload. As the dynamics of the system changes, newer data replace the old data, according to a custom data retention policy. The knowledge database increases the retention of data likely to boost the fitness of the solution and deprecates the data less likely to improve the existing solution.

Control parameters: Control parameters define the optimal decision boundaries that result in placement of the workload on the selected compute node.

Swarm agents: Swarm agents participate in the system optimization process by executing specific roles in a decentralized and self-organized system. These agents coordinate with each other and with the environment, ultimately leading to the emergence of intelligent global behavior.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.