In recent years,there has been a growing need for complex event processing (CEP),ranging from supply chain management to security monitoring.In many scenarios events are generated in different sources but arrive at the central server out of order,due to the differences of network latencies.Most state-of-the-art techniques process out-of-order events by buffering the events until the total event order within a specified range can be guaranteed.Their main problems are leading to increasing response time and reducing system throughput.This paper aims to build a high performance out-oforder event processing mechanism,which can match events as soon as they arrive instead of buffering them till all arrive.A suffix-automaton-based event matching algorithm is proposed to speed up query processing,and a confidence-based accuracy evaluation is proposed to control the query result quality.The performance of our approach is evaluated through detailed accuracy and response time analysis.As experimental results show,our approach can obviously speed up the query matching time and produce reasonable query results. In recent years, there has been a growing need for complex event processing (CEP), ranging from supply chain management to security monitoring. Many scenarios events are generated in different sources but arrive at the central server out of order, due to the differences of network latencies. Most state-of-the-art techniques process out-of-order events by buffering the events until the total event order within a specified range can be guaranteed. The major problems are leading to increasing response time and reducing system throughput .This paper aims to build a high performance out-oforder event processing mechanism, which can match events as soon as they arrive instead of buffering them till all arrive. A suffix-automaton-based event matching algorithm is proposed to speed up query processing, and a confidence-based accuracy evaluation is proposed to control the query result quality. The performance of our approach is evaluated as detailed accuracy and response time analysis. As experime ntal results show, our approach can obviously speed up the query matching time and produce reasonable query results.