Study Of Real-Time Performance In Embedded System

With the development of the embedded technology, the embedded system got an extensive application. Particularly in real-time monitoring fields, people often connect the embedded system with the computer network in order to make up the real-time monitoring network . In this kind of real-time monitoring network, the real-time of data communication is the most important problem. Because of the limit of the resources of the embedded equipment, it can not depend on its performance to raise real-time like the common PC. The traditional network technique has put great efforts to guarantee the realiability of data transmission and reduced the real-time of data transmission. So it can not be applied to the embedded monitoring system. How to improve the existing technique which can be suitable for the embedded monitoring system becomes a hot issue.The main contents of this thesis are included as follows:Firstly, this thesis presents the multi-processors framework which be used in the embedded system. With the development of the integrated circuit technique, the function of processor has already developed to a comparatively higher height. Raising the real-time of system just depends on raising the function of processor is not enough. So this thesis presents a new framework which uses multi-processors for working cooperation. When the data arrives the data-distribute-processor will distribute the data which has got to the data-disposal-processors. The multi-data-disposal-processors are symmetrical which can equally dispose the data. This framework can raise the real-time of system.Secondly, this thesis presents a new improved TCP congestion control algorithms. The traditional TCP congestion control algorithms has put great efforts to guarantee the reliability of data transmission and control the data flow, and it does not consider much for real-time system. So it is not suitable for the embedded system. By analyzing various existing congestion control we propose an improved algorithm. In this algorithm, we set the data packets different priorities which depend on the packet's importance, length and waiting time. The data packet which be set the priority will queue up for disposing. At the same time, the queue will be set three threshold values. The congestion happened when the queue arrives the first threshold value, the algorithms will discard the data packets which have the low priority. If the speed of discarded data is lower than the speed of arrived data, the queue will be longer. When the queue arrives the second threshold value, the algorithms will discard the data packets which have the normal priority. When the queue arrives the third threshold value, the algorithms will discard the data packets which have the high priority. Simulation studies show that the improved algorithm gives a better real-time compared to the traditional congestion control algorithms.Finally, this thesis presents a data steam query optimization algorithm based on embedded system. The collecting of data in the embedded system is returned by the steam form. Querying the data steam in time is the important method which improves the real-time of system. This thesis presents a new data steam query optimization algorithm which based on shared sliding window of two-dimensional table. It does a two-dimensional table which depends on the importance of query and the referring time of query. Every time the system does the query which has the max right in the two-dimensional table. Simulation studies show that the improved algorithm assures first excuting important query and consider the query which needs waiting for long time.