Year: 2012
Type: dissertation
Abstract: This thesis proposes Hierarchical Agent-based Adaptation (H2A), a scalable platform-based design paradigm for the rapidly expanding parallel embedded systems.With the constant progress of semiconductor industry, the available hardware resources are steadily increasing, while the constraints in power/energy consumption and dependability issues pose tougher challenges than ever before.This phenomenon calls for a system-level paradigm shift towards self-aware and adaptive (SAA) systems, which are able to provide dynamic performance optimization based on self-monitored status and incidents.The thesis motivates the paradigm shift towards SAA systems by examining the technology trends in parallel computing and embedded systems.Then it explores the existing techniques for run-time monitoring and reconfiguration, either with centralized, clustered or purely distributed architectures.The design and integration of these existing techniques and potential mechanisms require a scalable management backbone to provide global and local services, in particular for energy management and dependable computing.Answering this need, the thesis introduces the H2A system architecture, which integrates self-adaptation controllers, Agents, for monitoring and reconfiguring systems on hierarchical levels.The thesis overviews the functional partition of agents for coarse-and fine-grained services, and formulates the software/hardware (SW/HW) co-synthesis guidelines to implement the agents in a scalable manner.To demonstrate the design of agents for run-time adaptation, the thesis first presents hierarchical energy management service.Energy-aware mapping is performed on the top-level platform agent.Cluster-level energy management is designed as clustered decision-making and distributed reconfiguration (CDDR), which is jointly contributed by both the cluster and cell agents.To support CDDR, hybrid monitoring of local traffic loads and application milei SW/HW Software/Hardware VLSI Very Large Scale Integration xviii
Source: UTUPub (University of Turku)
Author: Guang, Liang
Language: English
Subfield: Hardware and Architecture
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2274195589