Refine
Document Type
- Doctoral thesis (2)
Has Fulltext
- yes (2)
Is part of the Bibliography
- no (2)
Language
- English (2)
Keywords
- Energieeffizienz (2) (remove)
Institute
Multi- and manycore processors promise to combine high overall peak performance with moderate power consumption to meet the constantly growing demand for computational power under the energy constraints of today’s CMOS technology. Future systems with manycore processors are expected to contain a huge amount of cores, which exceeds the number of processes that will run simultaneously. Consequently, processor time sharing approaches, that introduce significant overhead from regular context switches in common OS, will no longer be necessary.
This work investigates mechanisms for scalable and energy-efficient spatial partitioning of multi- and manycore processor systems. In addition, it explores the implications of exclusive processor core allocation to user processes due to the absence of temporal multiplexing and offers approaches to ease the adaptation to the new programming model.
The proposed mechanisms achieved fast thread allocation which motivates applications for dynamic thread allocation and benefits performance as well as energy efficiency. The efficiency control and resource revocation mechanisms detect and prevent wasteful and inefficient resource occupation from poorly optimized or malicious processes. In this way, the global efficiency of the system is optimized. The dynamic processing resource allocation and revocation handling has been integrated into a task parallel runtime system, to disburden the application programmer from manual implementation and to increase productivity.
This thesis deals with strategies for achieving high data throughput in wireless sensor networks that use a time division multiple access (TDMA) scheme to resolve medium access. The thesis uses a multi-sided approach that deals not only with the scheduling algorithm but also with the network layer and the interference model. The thesis proposes four solutions that significantly improve data throughput, fairness, and latency in the considered scenario.
The thesis starts with an overview of state-of-the-art medium access control (MAC) protocols, emphasizing TDMA. Based on this overview, it is concluded that not much space for improvement is left in the field of schedule calculation algorithms; many such algorithms are proposed up to date, and they can achieve schedule lengths close to the theoretical minimum. However, the research also reveals a lack of in-detail evaluation and comparison of these algorithms; this makes choosing the most suitable algorithm for a particular application hard and performance estimation inaccurate. Therefore, an extensive evaluation of state-of-the-art TDMA protocols using simulations and over 200 randomly generated networks was performed to tackle this issue. The results allow choosing an appropriate algorithm and estimating performance for each specific application.
Next, the problem of multiple packet transmissions during a single time slot is analyzed. State-of-the-art TDMA protocols assume that one packet can be transmitted in each slot and optimize the number of slots each node gets under this assumption. However, when nodes can transmit more than one packet, the performance of such a schedule is impaired. To solve this, the M-TreeMAC protocol is proposed; this protocol considers the actual number of packets transmitted in a time slot and optimizes the schedule accordingly. Furthermore, it is observed that the routing topology heavily impacts the schedule length created using this algorithm; an algorithm that optimizes the topology to result in the shortest schedule when M-TreeMAC is used is proposed, increasing benefits even further.
Finally, the accuracy of the 2-hop interference model, commonly used by state-of-the-art TDMA scheduling algorithms, is studied and simulated using a realistic radio model based on measurement results. The results show high packet loss ratios for packets traveling a large number of hops to reach the sink. The adaptive interference model is proposed to improve the 2-hop interference model. The proposed model can increase throughput significantly in networks with a height of ten or more hops.