Back
Texte intégral
Doctoral thesis

High-level services for networks-on-chip

    21.09.2012

170 p

Thèse de doctorat: Università della Svizzera italiana, 2012

Networks-on-chip
Network interfaces
Multiprocessors systems-on-chip
Security
Monitoring
Fault tolerance
High-level services
English Future technology trends envision that next-generation Multiprocessors Systems-on- Chip (MPSoCs) will be composed of a combination of a large number of processing and storage elements interconnected by complex communication architectures. Communication and interconnection between these basic blocks play a role of crucial importance when the number of these elements increases. Enabling reliable communication channels between cores becomes therefore a challenge for system designers. Networks-on-Chip (NoCs) appeared as a strategy for connecting and managing the communication between several design elements and IP blocks, as required in complex Systems-on-Chip (SoCs). The topic can be considered as a multidisciplinary synthesis of multiprocessing, parallel computing, networking, and on- chip communication domains. Networks-on-Chip, in addition to standard communication services, can be employed for providing support for the implementation of system-level services. This dissertation will demonstrate how high-level services can be added to an MPSoC platform by embedding appropriate hardware/software support in the network interfaces (NIs) of the NoC. In this dissertation, the implementation of innovative modules acting in parallel with protocol translation and data transmission in NIs is proposed and evaluated. The modules can support the execution of the high-level services in the NoC at a relatively low cost in terms of area and energy consumption. Three types of services will be addressed and discussed: security, monitoring, and fault tolerance. With respect to the security aspect, this dissertation will discuss the implementation of an innovative data protection mechanism for detecting and preventing illegal accesses to protected memory blocks and/or memory mapped peripherals. The second aspect will be addressed by proposing the implementation of a monitoring system based on programmable multipurpose monitoring probes aimed at detecting NoC internal events and run-time characteristics. As last topic, new architectural solutions for the design of fault tolerant network interfaces will be presented and discussed.
Language
  • English
Classification
Computer science and technology
License
License undefined
Identifiers
Persistent URL
https://n2t.net/ark:/12658/srd1318659
Statistics
Document views: 73 File downloads:
  • 2012INFO010.pdf: 81