As video surveillance images become more and more clear, storage time becomes longer and longer, and storage reliability requirements become higher and higher, video data storage gradually shifts from digital video recorders (DVRs) to professional IP SAN professional storage devices. The performance in the park monitoring is more obvious.
However, in the monitoring of the park, not only the monitoring of the security of the park, but also the monitoring of the production business is required. In the traditional monitoring, the media server file storage mode has not been able to solve the needs of new storage services, such as the "dumb-bell" effect of media servers, the rapid retrieval of production monitoring emergency, and the problem of historical video playback.
Solving the above problems requires rethinking the selection of storage modes in monitoring and breaking the file storage mode of media streams in traditional monitoring.
The traditional monitoring of media stream file storage uses streaming data to form a file for storage. All historical image playback is accomplished by playing back the stored video file. During file formation, management personnel cannot read the file, such as many “DVRsâ€. In the monitoring application of the "DVS+ streaming media server" method, the monitoring image is stored in a manner of forming a video file every half hour, which means that the historical image within half an hour before the current moment cannot be called.
In the event of an emergency, people want to be able to see historical images as soon as possible, and even require that they can be played back at any time. This requires that the time interval for the data stream to form a file is as short as possible, but this will cause a huge performance pressure on the system. If a video file is formed in 5 minutes, each monitoring point will generate nearly 300 files a day. If the historical image of each monitoring point is to be stored for 30 days, it means that each monitoring point needs to generate nearly 9000 files. A large-scale monitoring system of 1000 roads will need to handle 9 million files, and the system will be overwhelmed by heavy responsibility.
In response, the industry began to try new data management methods as early as a few years ago, hoping to improve the data management efficiency of monitoring storage and meet the needs of monitoring large-scale applications. Among them, "block storage" is a kind of characteristic one. This method adopts a special data structure of "time index + block data", abandons the traditional file system, and improves the management efficiency of monitoring data.
Performance bottleneck and reliability of media server In the traditional mode, when it is necessary to achieve centralized storage or support multiple people to watch a monitoring point at the same time, the media server is used for transferring or copying. In this way, the media server needs to convert the media stream into files for storage. The front-end media stream data is transferred to the back-end storage device via the media server. The processing performance and bandwidth of the server can easily become the performance bottleneck. The performance of the entire system is distributed into a dumbbell shape, which also increases the failure points. The failure of the media server affects the entire system's monitoring and storage services.
Media server file storage mode In fact, in many large-scale monitoring solutions, in order to solve the problem of media server performance bottleneck, the general use of server groups to complete. But it also brings new problems, such as how to load balance between multiple servers? How does the system switch data traffic to another server after a server failure? How are these servers managed? How to share a storage space? And so on, to solve these problems requires a very good cluster management system. While increasing the complexity of the system, it also requires a huge budget. What is more regrettable is that there is not yet a cluster management system in the industry that can solve this problem. problem.
Therefore, the end-to-end direct storage of the front-end device to the IP SAN is a good solution.
In the storage mode, the "data block direct storage" data management method abandons the media server. On the basis of the IP network, the iSCSI module is integrated in the encoding device so that the encoding device can end-to-end recording data based on the iSCSI protocol. Write to the IPSAN storage device.
Search efficiency of surveillance video In traditional surveillance, for the media stream file storage mode, the corresponding file is first searched according to the camera and retrieval time in the video retrieval, and then the specific time point is further located, and the video is played back from the time point. . The smallest unit of historical data retrieval is the file, with too large a granularity and low accuracy.
The file system is originally designed for data management applications that are read and written at random, and the retrieval efficiency is low. The retrieval efficiency of a system with several million files is hard to imagine. Therefore, when the scale of the system is increased, the efficiency of the traditional digital monitoring scheme will decline rapidly.
In a "block-by-block" system, "block" storage can be understood as a custom file system where data is read and written on a bare disk; time is used as an index for each data unit, and index and data units are saved In a completely independent logical storage space. The index and data of the video form an independent and complete data structure. This data structure is entirely managed by itself and is no longer managed by the operating system and files. With the combination of time index + block data storage, retrieval can be performed based on time in video retrieval, and the video can be quickly located to an arbitrary time, and the retrieval efficiency is greatly improved. The granularity of the search is no longer limited by the size of the file, and continuous search in seconds can be achieved.
Of course, this storage method will inevitably impose higher requirements on the technology accumulation and strength of the monitoring vendors. For example, the encoder is required to support dual stream, and the real-time monitoring video stream and storage video stream can be encoded independently, and the encoder supports the iSCSI protocol. Dual-stream is the basic requirement of all high-end encoders nowadays, and there are many vendors that implement it. Encoder support for iSCSI requires the monitoring vendors to have professional accumulation of storage technologies. For most current monitoring vendors, there are certain technical thresholds. Fortunately, we see that more and more monitoring vendors recognize the importance of storage for a monitoring program and begin to accumulate technology in this area.
However, in the monitoring of the park, not only the monitoring of the security of the park, but also the monitoring of the production business is required. In the traditional monitoring, the media server file storage mode has not been able to solve the needs of new storage services, such as the "dumb-bell" effect of media servers, the rapid retrieval of production monitoring emergency, and the problem of historical video playback.
Solving the above problems requires rethinking the selection of storage modes in monitoring and breaking the file storage mode of media streams in traditional monitoring.
The traditional monitoring of media stream file storage uses streaming data to form a file for storage. All historical image playback is accomplished by playing back the stored video file. During file formation, management personnel cannot read the file, such as many “DVRsâ€. In the monitoring application of the "DVS+ streaming media server" method, the monitoring image is stored in a manner of forming a video file every half hour, which means that the historical image within half an hour before the current moment cannot be called.
In the event of an emergency, people want to be able to see historical images as soon as possible, and even require that they can be played back at any time. This requires that the time interval for the data stream to form a file is as short as possible, but this will cause a huge performance pressure on the system. If a video file is formed in 5 minutes, each monitoring point will generate nearly 300 files a day. If the historical image of each monitoring point is to be stored for 30 days, it means that each monitoring point needs to generate nearly 9000 files. A large-scale monitoring system of 1000 roads will need to handle 9 million files, and the system will be overwhelmed by heavy responsibility.
In response, the industry began to try new data management methods as early as a few years ago, hoping to improve the data management efficiency of monitoring storage and meet the needs of monitoring large-scale applications. Among them, "block storage" is a kind of characteristic one. This method adopts a special data structure of "time index + block data", abandons the traditional file system, and improves the management efficiency of monitoring data.
Performance bottleneck and reliability of media server In the traditional mode, when it is necessary to achieve centralized storage or support multiple people to watch a monitoring point at the same time, the media server is used for transferring or copying. In this way, the media server needs to convert the media stream into files for storage. The front-end media stream data is transferred to the back-end storage device via the media server. The processing performance and bandwidth of the server can easily become the performance bottleneck. The performance of the entire system is distributed into a dumbbell shape, which also increases the failure points. The failure of the media server affects the entire system's monitoring and storage services.
Media server file storage mode In fact, in many large-scale monitoring solutions, in order to solve the problem of media server performance bottleneck, the general use of server groups to complete. But it also brings new problems, such as how to load balance between multiple servers? How does the system switch data traffic to another server after a server failure? How are these servers managed? How to share a storage space? And so on, to solve these problems requires a very good cluster management system. While increasing the complexity of the system, it also requires a huge budget. What is more regrettable is that there is not yet a cluster management system in the industry that can solve this problem. problem.
Therefore, the end-to-end direct storage of the front-end device to the IP SAN is a good solution.
In the storage mode, the "data block direct storage" data management method abandons the media server. On the basis of the IP network, the iSCSI module is integrated in the encoding device so that the encoding device can end-to-end recording data based on the iSCSI protocol. Write to the IPSAN storage device.
Search efficiency of surveillance video In traditional surveillance, for the media stream file storage mode, the corresponding file is first searched according to the camera and retrieval time in the video retrieval, and then the specific time point is further located, and the video is played back from the time point. . The smallest unit of historical data retrieval is the file, with too large a granularity and low accuracy.
The file system is originally designed for data management applications that are read and written at random, and the retrieval efficiency is low. The retrieval efficiency of a system with several million files is hard to imagine. Therefore, when the scale of the system is increased, the efficiency of the traditional digital monitoring scheme will decline rapidly.
In a "block-by-block" system, "block" storage can be understood as a custom file system where data is read and written on a bare disk; time is used as an index for each data unit, and index and data units are saved In a completely independent logical storage space. The index and data of the video form an independent and complete data structure. This data structure is entirely managed by itself and is no longer managed by the operating system and files. With the combination of time index + block data storage, retrieval can be performed based on time in video retrieval, and the video can be quickly located to an arbitrary time, and the retrieval efficiency is greatly improved. The granularity of the search is no longer limited by the size of the file, and continuous search in seconds can be achieved.
Of course, this storage method will inevitably impose higher requirements on the technology accumulation and strength of the monitoring vendors. For example, the encoder is required to support dual stream, and the real-time monitoring video stream and storage video stream can be encoded independently, and the encoder supports the iSCSI protocol. Dual-stream is the basic requirement of all high-end encoders nowadays, and there are many vendors that implement it. Encoder support for iSCSI requires the monitoring vendors to have professional accumulation of storage technologies. For most current monitoring vendors, there are certain technical thresholds. Fortunately, we see that more and more monitoring vendors recognize the importance of storage for a monitoring program and begin to accumulate technology in this area.
Others Electroplated Diamond Tools
Sintered Saw Blade Co., Ltd. , http://www.nsdiamondtool.com