Distributed Real-Time Transaction Processing

点击翻译按钮获取中文摘要

Many real-world applications such as airline reservations and stock markets require real-time response while interacting with large, geographically distributed, databases. Transactions in such a system have deadlines, but there are no catastrophic events if a deadline is missed; however, a penalty must be paid for late completion (same concept as paying your taxes after the April 30 deadline). The primary goal is to meet the deadlines, but for transactions that miss it, the delay must be minimized. For instance, a typical requirement may be that 95% of transactions complete in less than 2 seconds. Prior to building large/expensive networks and distributed systems, efficient protocols must be designed and tested to demonstrate their optimal performance under a variety of conditions. To do this cost-effectively, we propose a modular approach for development and testing in a simulated environment. Among other things, the simulator should be capable of simulating low, moderate and high load conditions. An interactive graphical user interface would allow a combination of input parameters to define various system configurations and data visualization. The sheer volume of data that is generated makes data analysis a very tedious task. We propose to apply business intelligence tools to scientific data for gaining insight into the results. For instance, predictive analytics may generate an early warning of possible congestion that could lead to poor system performance. This information can then be used to prescribe remedial actions to avoid performance degradation. This concept is similar to predicting traffic congestion several minutes before it is going to happen and then taking corrective action to avoid the imminent congestion. Business intelligence concepts have never been applied to data of this nature. The proposed research has potential applications in areas such as automated manufacturing, multimedia databases, real-time process control, banking, airline reservations, and stock exchanges. The analytics and data visualization modules can also be used in other applications which generate large volumes of data.