Due to the impact of the epidemic, most employees of enterprises cannot return to office buildings at present, and students postpone the start of school. Stable and efficient telecommuting and live teaching have become the starting point of 2020. In February, Tencent announced that it would open a free Tencent conference that could support 300 online meetings during the pandemic, which was also reported by CCTV news.
Click to see the video
Born near tencent meeting instantly became a national assembly software, massive users into the outbreak of geometric level, January 29th to February 6, tencent meeting every day in the resource capacity, average daily expansion cloud hosting close to 15000 units, 8 days a total capacity of more than 100000, cloud hosts involved millions of nucleus of computing resources, on February 10, Tencent conference backend server requests increased by 5 times.
So how does Tencent meeting achieve no perception during the expansion of users, still can hd smooth without the delay of the meeting? That has to mention a great soldier — Tencent Cloud Redis database (TencentDB for Redis).
In the case of a sudden increase in the number of requests, Redis database as a conference list cache, conference information storage has also been rapidly expanded, Tencent cloud Redis database by providing large-scale cluster products, Tencent conference to provide stable cache services with high concurrency and low delay, the operation is also extremely simple. The operation and maintenance personnel of Tencent Conference only need to click a button on the console to complete the elastic expansion of several times of business specification growth.
Ten million QPS Redis single cluster access performance
In order to respond to massive user requests, in the architecture of million-core computing business, the introduction of caching system is an effective method to guarantee the rapid growth of business scale while providing stable performance and fast response.
In large-scale user scenarios, Tencent Conference chooses Tencent Cloud Redis as a cache service to store the content of scheduled meetings and information of participants in Redis, ensuring that the system can respond quickly to users in the process of booking, initiating and participating in meetings. Redis service with Tencent clustering architecture can provide a maximum storage capacity of 4TB and concurrent access performance of 100,000-10 million level only in a single cluster, and can guarantee 1ms response delay within 99.99% water line.
Tencent Cloud Redis single cluster of more than 1.7 billion peak requests per minute
Within 30 minutes, it will expand tens of times without stopping
While Tencent conference completed the expansion of 1 million core cloud server in 8 days, Redis cluster efficiently completed the expansion of dozens of times in only half an hour. The background processing time of the expansion process of a single cluster did not exceed 30 minutes, while maintaining 100% system availability. In the whole process of resource expansion, Tencent conference service has always maintained a large-scale online operation, a large number of users have no perception, but still can high-definition smooth without delays.
It is worth being proud that Tencent Cloud Redis is the only Redis database product with non-destructive capacity expansion in China.
So what is the underlying design that allows Redis services to scale so smoothly and flexibly?
Tencent cloud Redis automatic resource management and packing system ensures the rapid allocation of resources, and provides standardized console interface and API interface, enabling operation and maintenance personnel to quickly expand resources. They only need to click a button on the console to complete the elastic expansion of several times of business specification growth.
In order to better provide services for users, Tencent Cloud Redis database takes the lead in realizing real-time smooth lossless elastic expansion in China. Most cloud vendor Redis database be disconnects when it is across the machine capacity and 1 minute business read-only, USES the self-built and other cloud vendors Redis service, enterprise can only rely on the business layer smooth extension, or stop taking to maintain the extension, the need to provide 7 * 24 hours customer service enterprise is fatal.
To achieve lossless capacity expansion, there are two core problems to be solved:
The first is to solve the community migration tool migration large Key blocking system access, or even downtime. Tencent Cloud team not only solved the problem of lag through self-developed data relocation tools, but also doubled the speed of data relocation.
Secondly, it is necessary to provide the correct marking data status in the process of data transport, so that re-routing problems can be solved after relocation. Tencent Cloud Redis periodically refreshes its self-developed Proxy and responds to route redirection commands to solve the problem of data routing.
Schematic diagram of Tencent Cloud Redis nondestructive capacity expansion
After years of rapid development, Tencent Cloud Redis has served many users in e-commerce, games and other industries, providing secure and stable cloud services for tens of thousands of users. Take an e-commerce customer as an example. Since its launch, it has rapidly accumulated hundreds of millions of users, and both UGC community business and B2C e-commerce business have developed rapidly. It is particularly important to improve users’ visiting experience and shopping experience efficiently. Tencent Cloud cloud cache Redis service requires no installation, one-click use, automatic expansion, easy processing and concurrent massive data, professional team monitoring, effectively help the customer to cope with the explosive growth of business scale and the smooth holding of various promotional activities.
Phase to recommend
With a maximum of 500,000 QPS, Tencent Cloud newly released Redis 4.0 standard edition breaks the performance limit
Preferential experience cloud database
Click on the offer to buy Tencent research database CynosDB