National Technical University of Athens (NTUA)
Ανάπτυξη συστηματικής μεθοδολογίας δυναμικής διαχείρισης πόρων σε ενσωματωμένα συστήματα
Abstract
dc:descriptionNext generation wireless systems support a wide range of communication protocols and services, thus opening new design challenges. The desired flexibility requires an effective utilization of system resources. A well-known example of a platform with these capabilities is Software Defined Radio (SDR). SDR terminals are critical to enable concrete and consecutive inter-working between fourth generation wireless access systems or communication modes. These platforms have heavy hardware requirements and switching between them introduce dynamism in respect to timing and size of resource requests. Very often these devices experience transient overloads due to workload bursts or hardware malfunctions. For that reason such systems have to be designed to take timely reactions to the occurrences of unexpected usage scenarios. In this thesis, we concentrate on these design challenges using the concept of the system scenario proposing solutions especially for wireless communication systems. We examine the special characteristics of the SDR wireless platforms concentrating on the baseband signal processing. Our work is not limited to applying system scenarios; we propose novel methodology extensions for the clustering of system behaviors into system scenarios and the run-time detection of these scenarios. More precisely, we study the main trade–off between clustering overhead (which is correlated with how representative are the scenarios), detection overhead (proposing transformations at the implementation of the detection algorithm) and switching overhead, which represent the tuning cost of the platform. In the context of our study, we exploit the flexibility of a system-level framework, combining a SDR coarse-grain simulator with a cycle-accurate Network-on-Chip simulation environment. For the needs of our analysis, we used an SDR application modelled as task graph. The simulation flow can be used at an early design phase. The scope is to enable a pre-implemented cycle accurate characterization of a potential SDR platform extracting the optimal system configurations. The automatic exploration of the SDR platform and the interface between the two layers succeed through two wrappers, which ensure the automatic execution of our framework, and thus the automatic exploration of different SDR hardware platform configurations. A set of cost trade-offs is defined and can be exploited to the final platform development. The aim was to be achieved an efficient resource utilization retaining the reconfiguration cost in reasonable levels. More precisely, at the first case study, using a WLAN communication protocol, we succeeded a reduction of the idle times 92% on average at the execution of each block of bits, in comparison with the worst case. In the second case study, it is showed that the exploitation of the system scenarios allowed 1) a significant decrease of the tuning situations (from 54 Pareto curves with 3888 design space points to 7 Pareto curves with 49 Pareto points) and 2) a remarkable energy saving about45-73% (based on the system’s deadlines). In the third case study, we apply a signal power scaling technique based on the running system scenario decreasing the antennas energy consumption 50-94%. The results are encouraging and show that system scenario methodology are applicable at dynamic wireless systems and can cope with their complexity.Finally in the last section, we propose a systematic methodology to predict the workload trend based on advanced mathematical tools from the Chaos Theory domain. The benefit of our approach is that we can analyse workloads with high complexity without requiring knowledge of the specifications of the target application. To evaluate the effectiveness of our approach we applied a DFS technique based in our workload analysis simulating a complex dynamic multithreaded application in a cycle-accurate NoC simulator. Using the proposed DFS strategy the designer can achieve remarkable power consumption improvements, ranging from 17.5% (hard timing constraints) up to 37.8% (relaxed timing constraints), depending on the timing constraints of the application.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2014
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Zompakis, Nikolaos
- Ζομπάκης, Νικόλαος
Subjects
dc:subject × 14- Σενάρια συστήματος
- Σύστημα
- Μεθοδολογία
- Ενσωματωμένα συστήματα
- Δυναμική διαχείριση πόρων
- System scenarios
- System
- Methodology
- Embedded systems
- Dynamic resource management
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Engineering and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
Rights
dc:rights- Statement dc:rights
-
- BY
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/39743