What is the core content of security cloud computing?

For the intelligent video image analysis service, the main core is the intelligent analysis algorithm engine. Currently, the industry is generally based on the deep learning-based intelligent video image analysis algorithm. At the IaaS infrastructure level, high-performance GPU server equipment is generally used to improve the algorithm execution. effectiveness. Therefore, on the security cloud platform, the GPU server device cluster has become a necessary resource pool, providing a high-performance computing foundation for the security cloud platform.

In response to this demand, Keda's self-developed intelligent analysis GPU server provides high-density and high-performance GPU computing capabilities. It can provide no less than 12 high-performance GPU analysis and processing units on a standard 4U server architecture, supporting stack expansion. It can fully meet the needs of various depth analysis intelligent analysis algorithms clustering and high-density applications.

At the IaaS layer system level, the security cloud computing platform needs to efficiently manage and utilize GPU computing resources, and adopts clustering and virtualization to flexibly schedule and allocate resources to maximize video image analysis capabilities. At the same time, the security cloud platform needs to be able to closely monitor the GPU computing resource usage and the state of the GPU processor. In the case of single or multiple GPU computing device failures, it can ensure that the intelligent video that is already in progress is not affected. Analyze the business and implement the GPU's pooled service performance. The built-in cluster management function of the GPU server used in the Keda intelligent analysis system can provide rich resource monitoring management and scheduling services in combination with the requirements of various video intelligent analysis services in the security cloud platform, and realize the efficient application of the cloud platform to the GPU computing resources.

In addition, at the PaaS level, the security cloud computing platform can further layer the intelligent video analysis service, separate the algorithm framework and specific algorithm applications, and provide various types of deep learning analysis algorithm frameworks, such as TensorFlow, Torch, etc. Ensure that these algorithm frameworks can quickly dispatch various computing resources with a powerful cloud infrastructure to provide a unified, reliable, and convenient resource service platform for specific algorithm applications. At that time, various specific deep analysis-based intelligent analysis applications, such as face analysis, vehicle analysis, and behavior analysis, can be quickly deployed and managed on the cloud platform.

For security big data analytics services, the main core is a variety of big data analytics mining algorithms (PaaS level), and a distributed database (DaaS level) that can support these data analysis algorithms. Security big data analysis is faced with massive security data resources, including billions of billions of tens of billions of target description information and related feature information, including people, vehicle target records, portrait features, face features, vehicle characteristics, and A huge vehicle information base, face information database, and so on. These data require highly reliable storage and read and write, and can be analyzed and utilized with high performance. At the DaaS level, Kodak uses high-performance distributed database technology to meet the needs of billions of data storage, recall and analysis through distributed computing and horizontal and convenient extension. Distributed database provides high-reliability, high-performance and easy-to-expand data storage service system for massive data by adopting data discrete storage, data redundancy protection, data tiered storage, memory acceleration, distributed computing and other technologies to meet the security cloud platform. Requirements for massively structured/semi-structured data storage, read and write, and analytical calculations.

On the basis of the distributed database, the security cloud platform deploys various distributed big data analysis and calculation engines on the PaaS level to analyze massive data resources, such as offline processing using MapReduce data, and implementing data near spark or strom. Real-time processing or real-time business processing. Engines that support big data full-text search, such as slor+spark, support data mining engines such as spark\Hive. Through these high-performance big data analysis algorithms and streaming data processing technologies, based on high-performance distributed databases, the efficiency of massive data analysis is further improved, and real-time monitoring alarm requirements can be met, thereby enabling post-processing to The shift in pre-warning. (Author: Su Jie)