Development of High Speed Data Acquisition System using Multiprocessor Array

近年来，随着能量的上升，高能物理实验的规模越来越大。数据采集与处理系统必须处理大数据量和复杂的数据格式。它不仅要求数据传输的容量更大，而且要求数据采集的方法更先进。除了对数据进行实时的校正和压缩外，还需要智能的数据预处理系统来处理海量数据。为此，利用INMOS的TRANSPUTER开发了多处理器系统。微处理器具有四个相互通信端口的设施；LINK提供了阵列中微型计算机节点之间的简单连接。为了在实验环境下对该系统进行评估，设计了一种阵列系统，应用于TRISTAN VENUS探测器中心漂移室命中线信息的寻迹处理。在本研究中，我们设计并构建了一个接口电路，更多地从CDC读出电子器件中获取命中线数据。并开发了阵列数据并行处理软件系统。该接口由通用控制部分和专用部分组成，该部分必须为其他应用程序重新设计。所测试的多处理器阵列由20个TRANSPUTER节点组成，其数量易于扩展。虽然分布式处理软件仍在改进中，但通过加载真实实验数据，对系统在当前版本软件下的性能进行了测试。结果表明，数据的平均处理时间约为几毫秒，而不是富士通M-780的20毫秒。结果表明，该系统的性能足以满足目前在VENUS实验中的应用。已经讨论了这一发展的报告。1990年4月9-13日在美国新墨西哥州圣达菲举行的“高能物理计算”会议上报告了上述结果。少

In the recent years, high energy physics experiments are getting the size bigger and bigger as the energy goes up. The data acquisition and processing system must deal with the large data size and complex data format. It requires advanced method for the data acquisition other than a larger capacity of the data transmission. An intelligent data preprocessing system is required to cope with the large data in addition to the data correction and compression in real time. Along this aim, a multi-processor system was developed using the TRANSPUTER of INMOS. The microprocessor has a facility of four communication ports for each other; LINK which provides an easy connection between microcomputer nodes in the array. For evaluation of such system in the experimental environment, an array system was designed to apply for the track finding processing based on the hit wire information of the central drift chamber of TRISTAN VENUS detector.In this research, we designed and built an interface circuit … More to squire the hit wire data from the CDC readout electronics. Also the software system for the parallel processing of the data in the array was developed. The interface is consist of the general control part and the dedicated part for this application which must be redesigned for the other application.The multiprocessor array under test consists of 20 TRANSPUTER nodes, the number of which is expandable easily. Although, an improvement of the distributed processing software is still under way, the performance of the system under the current version of the software were measured by loading the real experimental data. The results show the average processing time for the data is about a few milliseconds instead of 20 msec by Fujitsu M-780. The result shows the performance is enough for the present application for the VENUS experiment.A report of this development has been discussed. in the symposium on the Next Generation System of Data Acquisition in March 10-11, 1989, And the result is also reported at the conference of Computing in High Energy Physics '90 held at Santa Fe,New Mexico, April 9-13 1990. Less