The current price hikes in storage products have been around for almost a year, and there is still no sign of stopgap in this state of complete market performance. In a recent Garter report, they said they expect to wait until 2019 if the price of storage devices is cut. The report shows that in addition to smart phones and PCs, the Internet of Things devices, autonomous driving and so on have a huge demand for memory chips, the gap is also great. While some memory chips may be the end of the year gains have declined, but the premise is the price has risen to a certain extent. As information technology continues to evolve, new storage technologies such as flash memory, disks, data centers, and DNA continue to emerge. Even so, it is still difficult to meet the growing storage needs of data volume, coupled with the development of IoT (Internet of Things) industry, resulting in a more alarming volume of data. It is undeniable that many of these data contain value, but can not ignore the alarming volume of data. Do we want to all 44ZB data records and stored by 2020? Therefore, we need to use cloud computing technology to intelligently analyze the data. Today to discuss 2017 security cloud computing core technology. Large-scale hybrid computing If only a large number of video image data generated by the monitoring system are processed manually, the efficiency will be very low. With the video intelligent processing algorithm, some simple features can be obtained from video image data for comparison or pattern matching Alarm events, improve the efficiency of processing. The amount of data, the degree of data composition, the type of data, etc. that can be handled in this way are still low and can not cope with the massive data and the ever-increasing demand. The purpose of large-scale computing technology is to provide a unified data processing platform, which integrates various intelligent algorithms and computing models to comprehensively process massive monitoring data to obtain more valuable data faster. Discussion: 2017 Security cloud computing core technology Uniform resource management technology The main data generated by the monitoring system is the video and image data. After the original data is processed, it will produce richer data and the way of processing will be greatly different. For example, historical video data can be processed in the background of the video data retrieval, license plate and face feature data for the bayonets need real-time cloth control, historical mount information needs to be done in real time retrieval. These data all need different computing frameworks to deal with. By introducing a unified resource management platform, different computing frameworks can be operated in the same resource pool to greatly improve the utilization of resources. At the same time, when resources are monopolized by a certain kind of business, Can maximize the performance of the system. Real-time retrieval technology The traditional structured data are stored in relational databases. Database clusters are formed by techniques such as RAC and accelerated by indexing. However, the core is still based on row storage and relational operations. In the face of massive records, they have encountered bottlenecks in all aspects . Real-time retrieval technology can deal with 100 billion levels of structured data by introducing technologies such as distributed database, columnar storage, memory computing, indexing engine and so on, which can greatly improve the storage capacity, scalability, retrieval speed and other aspects . The system has important research value and broad application prospect in the field of video surveillance such as intelligent transportation and criminal investigation. Complex event processing techniques With the development of the security industry, the business becomes more and more complicated. For example, in the field of intelligent transportation, the demand for vehicle points research, deck car analysis and same-car analysis has emerged. These requirements exist to produce the results depend on many conditions, the process of real-time requirements of high, the need to deal with a huge amount of data and so on. The traditional way is to use a relational database, through the combination of complex SQL statements, constantly check the way of comparison, it is difficult to meet the real-time requirements. Complex event processing By introducing streaming computing and other technologies, dynamic analysis of input data in real time, the processing speed can be provided substantially. Not meet the conditions of the data are discarded, the system only exists in the processing results or may be useful intermediate data, so the requirements of the storage becomes smaller, completely in memory for the whole process of analysis, real-time be guaranteed. Face search technology Face retrieval technology in a single server application has been more mature, can be used in identification, fugitives arrested, suspicious staff to investigate, ID check and other fields. Face detection process can be divided into the following stages: video or image decoding, face detection, feature extraction, feature comparison, the first three steps are each request corresponds to a calculation, the amount of control is relatively controllable, and the last one Compared with each request, the step feature needs to compare with the facial features of hundreds of millions of levels, which is a stage with the largest amount of computation. Some real-time application requests up to hundreds of requests per second up to millions of face alignment, the entire system needs to support 100 million face features per second comparison. Such a large-scale calculation, stand-alone can not be completed, the cluster must be completed. The feature library itself is small in size but large in comparison, which belongs to a typical compute-intensive cluster. The feature library can be all poured into memory and completed in memory. Massive video retrieval technology After the video data collected by the image sensor is saved to the back-end storage, the user can select multiple cameras in the target area at any time and submit the video data to the video retrieval cluster. All clusters corresponding to the characteristics of the target object are quickly searched to generate video data and find the target The appearance of the object features the video, and locate the exact point in time. Which mainly uses the intelligent technology to achieve the conversion of video data to object feature structured data, support for vehicle color, license plate, clothing color, face and other features. Based on a unified computing resource pool, parallel computing of intelligent algorithms can be implemented to improve the retrieval efficiency linearly.