Baidu Builds Intel Optane-Based Memory Platform
| Intel Optane DC Persistent Memory and DRAM
With a hybrid memory configuration design that incorporates all components
| Improved 'Feed Stream' to provide customer-tailored content

Intel's memory capabilities are reflected in Baidu's internet services.

Intel announced on the 2nd that Baidu is designing an in-memory database for its 'Feed Stream' service to utilize the large-capacity, high-performance capabilities of 'Intel Optane DC persistent memory'.

By building a memory platform equipped with 2nd generation Intel Xeon Scalable processors and based on Intel Optane DC Persistent Memory, Baidu has been able to provide personalized search results to users while lowering the Total Cost of Ownership (TCO).

Intel and Baidu [discussed] details regarding this cooperation between the two companies at the 2019 Baidu ABC Summit held at the Beijing National Convention Center on August 29.Revealed.

Internet service providers like Baidu have been forced to achieve fast and efficient data access and storage as they manage explosively growing data. With this announcement, Baidu is improving its feed stream services to provide more personalized content to customers.

Baidu uses the in-memory database 'Feed-Cube' to accelerate data storage and information retrieval in its cloud-based feed stream service. By deploying Intel Optane DC Persistent Memory and 2nd Gen Intel Xeon processors, Baidu ensures concurrency and high performance for large volumes for Feed-Cube.

Intel and Baidu designed a hybrid memory configuration within the Baidu Feed Stream service that incorporates both Intel Optane DC Persistent Memory and DRAM. Through this approach, FeedCube improved response times for search results even under the pressure of massive concurrent traffic. At the same time, single-server DRAM usage was reduced to less than half, and costs were reduced in terms of FeedCube's petabyte-scale storage capacity.

Intel and Baidu released case studies on the research, including examples of other applications using Intel Optane DC Persistent Memory technology, such as Redis, Spark, and service functions.

Jason Grebe, Vice President and General Manager of Intel’s Cloud Platform and Technology Group, said, “Intel has been supporting Baidu’s core businesses, ranging from search, AI, and autonomous driving to the cloud, for over 10 years,” adding, “Through this close collaboration between the two companies, we are rapidly deploying the latest Intel technologies and enhancing the customer experience of Baidu’s services.”

Tao Wang, Senior Architect of Recommendation Technology Architecture at Baidu, stated, “Using Intel Optane DC Persistent Memory for the FeedCube database allows for cost-effective scaling of memory capacity to meet the demands of feed stream services.”

In addition, Intel and Baidu unveiled a storage solution that accelerates machine learning performance in high-performance computing (HPC) workloads. The new HPC solution is delivered within Baidu's cloud environment and provides users with end-to-end HPC support encompassing a wide range of functions, from data preprocessing, model training, and evaluation to inference and result publication. Baidu Cloud ABC Storage Service uses Intel Optane DC SSDs and Intel QLC 3D NAND SSDs.

Meanwhile, Intel and Baidu recently joined the Confidential Computing Consortium (CCC) under the Linux Foundation. As part of the consortium, Intel and Baidu plan to collaborate with industry partners to build and store private trusted computing services on the Baidu Cloud based on Intel Software Guard Extension (SGX) technology.