The end of May in Guangzhou, China is hot and humid, but it does not dampen attendee enthusiasm at the Tencent Cloud Summit. The summit’s main theme is Artificial Intelligence, and Intel is proud to be a key partner with Tencent to deliver AI innovations from cloud to edge.
Tencent has been a front runner of the desktop and mobile Internet eras, and is well positioned to thrive in the data era that is unfolding in front of us. As of now, Tencent’s WeChat has exceeded more than 1 billion monthly active users (MAU), and Tencent QQ still maintains close to 800 million MAU. WeChat, sometimes referred to as a super app, is also a platform that allows many partner apps to run on top of it. You can make voice or video calls, hail a cab, transfer money, purchase train or plane tickets, make reservations, order takeout, and do many other things without even leaving the app. Sitting on a wealth of data, Tencent is investing in AI technologies to enhance the user interface, drive actionable insights, and in turn enable industry digital transformation through AI and analytics offerings in Tencent Cloud.
Joint AI inference workload performance optimization
Artificial intelligence, similar to human intelligence, can roughly happen in two stages: Training and inference. Just like a kid who needs to go to school to acquire skills, a neural network needs to be trained to understand the patterns in data, structured or unstructured.
After a neural network is trained, it is deployed to run inference—to classify, recognize, and process new inputs, just like a trained worker applying his or her skills to get the job done. While AI training is often done in a cluster of compute nodes, AI inference needs to happen everywhere an application runs, in the data center, at the network edge, or on smart devices. Intel has a broad portfolio of products that enables efficient AI inference from cloud to edge.
Let’s take WeChat as an example. Voice recognition is becoming an important way to enhance the WeChat user interface and convert user voice into text for hands-free input. In a panel discussion at the summit, Tencent CEO Pony Ma discussed the importance of complete voice interaction design for Wechat to be used safely in cars in the future. Intel worked side-by-side with Tencent’s WeChat team to optimize their voice recognition workload through improving the Long-Short-Term Memory (LSTM) model on the latest Intel® Xeon® Scalable processor, achieving 16X performance boost according to Tencent, and a much more streamlined user experience of voice to text conversion.
Ubiquitous computing, from cloud to edge
To support a series of emerging workloads, such as the Internet of Things (IoT) and autonomous driving, cloud computing is going through a fundamental shift from highly centralized hyperscale data centers to more distributed, decentralized architectures, where computing resources are located at the edge, much closer to where the data is being generated and needs to be crunched. By extending cloud to the edge, latency and expensive data movement issues can both be solved.
As Tencent Cloud increasingly focuses on industry solution, deploying IoT sensors and computer vision engines at the edge become additional building blocks for smart retail or public security projects. Aside from Intel® Xeon® Scalable processors with optimized AI frameworks running in the cloud, Intel® Movidius™ is an ultra-low power computer vision engine realizing AI inference capabilities in gateways and devices. Tencent Cloud is collaborating with Intel around Movidius in a number of industry engagements to extend computing resources to the edge.
Creating New and Differentiated Cloud Services
Leveraging Tencent’s worldwide presence and edge computing services, Intel supported delivery of Tencent Energy Cloud, the first energy sector-specific cloud computing solution with solutions provider Gao Energy. This is a highly vertically-oriented package that provides a secure, reliable, efficient, scalable cloud infrastructure-in-a-box, combined with a smart data management platform, that energy sector customers can deploy at edge locations with intermittent Internet access. It is powered by Intel® Xeon® Scalable processor and the Gao Energy router running on the Intel® Atom® processor.
Last but not least, Tencent has been collaborating with Intel for the last few months to perform testing and validation on engineering samples of Intel® OptaneTM DC persistent memory, and has seen improvements in throughput, durability and quality of service. Intel persistent memory and the new generation of Intel® Xeon® Scalable processors combined are expected to better service highly concurrent applications such as e-commerce and the ever popular giving of the Chinese Spring Festival red packets.
We are thrilled that Tencent plans to offer Intel Optane DC persistent memory based cloud storage and database services, and share the benefits of Intel’s new innovations with their customer base. Intel will continue to work with our cloud service provider and other ecosystem partners to bring new generations of cloud product solutions to market.
For more information about Intel’s latest products for cloud builders, and joint innovations with our cloud service customers, please go to www.intel.com/CSP.