Home > News content

Star sea SA2 cloud server helps Tencent's advertising "thousands of people, thousands of faces", and the speed of advertising calculation is increased by 25%

via:博客园     time:2020/4/17 22:31:23     readed:647

Recently, Tencent's advertising related technical director revealed that Tencent's advertising retrieval real-time computing business supports tens of billions of advertising retrieval computing tasks every day with the help of the stable and high-performance computing power provided by Tencent cloud self-developed star sea SA2 cloud server. Compared with traditional servers, the computing time to match an advertisement is reduced 25%, effectively improving the accuracy of Tencent advertising recommendation.

Based on the user group's interest label, recommend the most suitable advertising information, and at the same time display it in the favorite creative form of the target audience

In fact,

In order to ensure that the presentation of advertising does not affect the product experience, the whole process needs to be completed in milliseconds, and the real-time requirements are very strict. In addition, because of the huge volume of business, Tencent advertising retrieval real-time computing system concurrency is very high, computing performance requirements are very large.

Tencent ad used a physical server cluster to deploy the service. The growing business also poses new challenges to the cost and flexibility of physical servers. Within Tencent

The underlying physical server of Xingxinghai SA2 cloud server adopts Tencent cloud's first server hardware with completely independent intellectual property rights.

In order to ensure that the star sea SA2 cloud server can effectively meet Tencent's business demands, Tencent advertising conducted multiple rounds of tests at the beginning of large-scale application. The test results show that compared with the traditional physical servers used in the past, the star sea SA2 cloud server reduces the overall advertising computing time by 25%.

In the test, Tencent also found that if it directly deployed the same physical server architecture in the past, the server performance advantage could not be highlighted. However, after virtualization, Xingxinghai SA2 cloud server, as an example on the cloud, can give full play to the performance advantage of the underlying hardware.

Under a physical environment, for example, CPU remote access to memory is present on a general server, increasing the computational time of CPU. But Tencent cloud self-developed virtualization technology, can be physical machine CPU and memory segmentation, to ensure that the entire single chip CPU allocated to the business, the business can always be close to the local memory.

In addition, in the scenario of ad computing, operating massive data will carry out frequent hard disk or memory read-write operations. The whole process has a great loss on the physical machine, and the physical machine fails, which is very easy to affect business availability. Thanks to the cloud fault monitoring capability, most of the security risks can be found immediately. With the help of the cloud computing operating system vstation independently developed by Tencent cloud, the cloud server hot migration is initiated automatically. The whole migration process has no perception of the business level, which ensures the business stability and continuity to the greatest extent.

Following up, Tencent Advertising plans to deploy more business on the Star Sea Cloud SA2 cloud server, and polish a better advertising service experience with the help of products and technologies on the cloud.

At present, Tencent cloud is accelerating the deployment of star sea SA2 cloud server in data centers all over the world after obtaining stable application verification in Tencent's internal business scenarios, opening stable and cost-effective computing capabilities to users in the form of cloud services.

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments