الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 145 |  |  |
| | | from | 145 | | |
Abstract
Modern programming languages are provided with a mechanism for automatic memory management called garbage collection. However, some problems existed when applying such languages to the real-time applications. Among these problems, there was the threat of predictability and schedulability of hard real-time tasks. As a solution, different concurrent garbage collection scheduling strategies had been presented. Unfortunately, these concurrent garbage collectors rely on a relatively large amount of redundant system memory. Therefore, some other effort had further been exerted. A latter algorithm that minimizes the worst-case system memory requirement with the schedulability of tasks not jeopardized is the deferrable server based garbage collector scheduling strategy. The deferrable server based garbage collection had been proven to surpass all other garbage collection scheduling strategies in minimizing the system memory requirement under a fixed worst-case garbage collection cycle time environment.The thesis role can be represented into two main points as follows: Firstly, the performance of the deferrable server based garbage collector is explored under a variable garbage collection cycle time environment. It also investigates the selection of thresholds to further reduce the system memory requirement under a fixed worst-case garbage collection cycle duration environment. This can be performed by using high capacities of the deferrable server task which were discarded before. Furthermore, the performance of using high server budgets is examined under a variable garbage collection cycle time environment. The simulation results demonstrate that the deferrable sever based garbage collector achieves better performance even under the actual real-time garbage collection environment than other garbage collection scheduling strategies. Moreover, simulating the system using the actual-case situation leads to results with higher CPU utilization. The simulation results also show that using high values for the server capacities can lead to better memory usage if a proper value of threshold is chosen even under the actual-case situation of garbage collection task.
Secondly, concurrent garbage collection based on a single aperiodic server such as the deferrable server or the sporadic server pays much attention due to its effectiveness in the real-time systems scene. However, the real-time environment taken for testing such algorithms involves a single aperiodic task; garbage collection task. In the case of having different types of traffic, with short deadlines and long deadlines, the single server provides poor performance. The garbage collection task may have to wait till a less urgent or a higher deadline request finishes its execution which leads to an increase in the system memory requirement and perhaps a deadline miss of garbage collection thread.
This thesis concentrates on minimizing the system memory requirement when there are multiple sources of events with garbage collection task by introducing a proposed concurrent garbage collector. In this collector, the system will have multiple servers; rather than one in other garbage collectors. These servers can either share or not share their capacities, i.e. a server can use the unused capacity of other servers in case of sharing. The two schemes give preference to higher priority servers. However, after investigating both strategies, the capacity sharing scheme gives lower response times for jobs with short deadlines, like garbage collection task, than without capacity sharing scheme. The simulation results show that multiple servers with capacity sharing based garbage collection scheduling strategy with all its variations exhibits better performance in terms of reducing the system memory requirement and meeting most of deadlines than the single server and multiple servers without capacity sharing based garbage collectors. Hence, embedded real-time systems no longer necessitate large areas of memory.