Maier S (2013)
Publication Language: English
Publication Type: Thesis
Publication year: 2013
With the rise of multi- and many-core processors, the challenge to utilize parallel computing power moves more and more into the focus of research. To benefit from many-core architectures, changes to the algorithms and data structures of the applications are needed. Furthermore, adjustments in the architecture of operating systems are necessary because they provide the underlying system services and functions to the application such as communication and synchronization of threads.
The aim of this thesis was to design and implement a process-based kernel, which consistently applies non-blocking synchronization to ensure the scalability on machines with a large number of processor cores. In contrast to most existing operating system kernels, this new kernel was designed from scratch with scalability in mind.
After its implementation, the performance relevant properties of the new kernel were evaluated using a 48-core x86-64 machine. The performance measurements are based on four specifically developed tests, which consist of parallel algorithms. For evaluation purposes the results were compared with those of two other kernels. Therefore, the common process-based Linux kernel and the event-based LAKE kernel for multi-core machines, which was developed at our chair, were used. The evaluations revealed that non-blocking algorithms in operating system kernels lead to promising scaling characteristics.
APA:
Maier, S. (2013). Leveraging Non-Blocking Synchronization in a Process-Based Many-Core OS Kernel (Master thesis).
MLA:
Maier, Sebastian. Leveraging Non-Blocking Synchronization in a Process-Based Many-Core OS Kernel. Master thesis, 2013.
BibTeX: Download