Yao Ting from Uniview Technology: How to deal with the storage challenges of security data under AloT
Under the AIoT, the data requirements for storage systems have changed, and the traditional storage architecture is no longer suitable for current storage and applications. Focusing on the requirements of capacity, performance, data value improvement, and operation and maintenance management, we need to achieve multi-device resource pooling, unified management and allocation of resources, and linear growth of capacity and performance.
On November 12, the “MTS2021 Storage Industry Trend Summit” hosted by TrendForce, an international high-tech industry research institution, was grandly held in Shenzhen. Yao Ting, deputy director of the cloud storage development department of Uniview Technology, shared the evolution and challenges of storage technology under AloT based on the company’s experience.
Traditional storage architecture no longer works
With the development of cloud computing, big data, Internet of Things, artificial intelligence and other businesses, data is growing exponentially. At the same time, there have been many changes in the form of data, from the original structured data to the current semi-structured data. Yao Ting first emphasized: “The traditional storage architecture is no longer suitable for mass data storage.”
For example, the original SAN and NAS storage is mainly vertical expansion, limited by performance and storage capacity bottlenecks, and the reliability of data is also limited to the device itself. With the increase of application systems, the maintenance and management of equipment will be relatively complicated, and resources and space may not be fully utilized.
Taking the security industry deeply cultivated by Uniview as an example, Yao Ting focused on the storage challenges brought by security data under AIoT. From the perspective of security data, in the mixed storage mode of various data, a large number of small files lead to the deterioration of ordinary storage performance; AI activates the original data, and the reading demand increases significantly, subverting the original model; the new data generated after analysis and cleaning, the value of Significant improvement requires higher reliability; how to achieve better TCO for the system has become an urgent issue.
How to achieve better TCO?
Yao Ting believes that to meet the current challenges, the following points should be achieved: multi-device resource pooling, unified resource management and allocation, linear growth of capacity and performance, to meet the storage requirements of high performance and massive data; block, file, and object storage Integrated services to meet diverse application needs; high-reliability data node protection, business failover, to ensure data reliability and business continuity.
Yao Ting pointed out that there are many new changes in the business of AloT, and issues such as how the system can achieve a better TCO (Total Cost of Ownership) has become an important concern of Uniview. Yao Ting believes that this requires the integration of storage and computing, business and intelligence.
Based on the grasp of industry demands, Uniview has launched a cloud storage hyper-converged solution. With the help of a hyper-converged architecture including external business layer, core processing layer and hardware device layer, unified operation and maintenance services can be provided.
Among them, the cloud storage solution can pool the resources of the original single device to achieve the purpose of resource sharing. Different from the traditional expansion mode, it adopts a fully symmetric distributed architecture and supports online expansion. As storage nodes increase, storage capacity, computing, and performance will increase linearly.
For the integration of business and computing, the storage solution is equipped with a computing board card, which can provide resources to the computing board card through storage, thereby pulling up the computing resource pool to achieve pooling. The computing board will provide virtual machines to provide business services for the upper layer.
In order to achieve intelligent integration, Uniview has also made many improvements to the storage solution, such as the deployment of dedicated high-density GPUs. According to Yao Ting, a single storage host can support up to 5 GPU boards, each GPU board has 2 dedicated GPU chips, and a single storage host can support up to 10 dedicated GPU chips.