White Paper

Intel MCA+MFP Helps JD Stable and Efficient Cloud Services "Improving the quality of the cloud host is very important to customers, which can greatly reduce business interruption. In order to achieve the 99.975% SLA commitment of the cloud host, JD and Intel has cooperated deeply, using Intel's MCA and MFP technologies to narrow the fault domain and perform fault isolation or online migration in advance, reducing the impact of memory CE and UE failures on the cloud host." —— Gong Yicheng Head of IaaS Product R&D of JD Cloud Business Group of JD Technology "System reliability and stability have always been the most basic service tenet of the services of JD Cloud, and JD Cloud has never stopped exploring basic technological breakthroughs. This time, JD Cloud and Intel have customized and optimized the Advanced RAS function of CPU, and jointly researched and developed the fault prediction model of Intel MFP in combination with usage scenarios to realize memory failure prediction and repair. This technology will effectively improve the reliability and stability of JD Cloud's services. We also expect to continuously improve the reliability and stability of the hardware system through the efforts of both parties. " —— Chen Guofeng Head of Hardware R&D of JD Cloud Business Group of JD Technology Contents Foreword......................................................................1 Background.................................................................1 Pain Point: The surge in business puts a higher requirement for the stability and reliability of JD Cloud .............................................2 Intel MCA Recovery + MFP helps JD Cloud provide efficient and stable services .......................................................................2 Conclusion ..................................................................5 Foreword "With the successful deployment of Intel’s MCA Recovery and MFP technologies, the down probability of JD Cloud’s host due to memory failures has been reduced by 40%, and the success rate of hot migration under memory failures has been increased by 50%, which has greatly enhanced the reliability and stability of the host, improved the user experience, and promoted the SLA, so that JD Cloud has a technological advantage in the fierce market competition." Today, with the rapid development of digital economy, the following life scenarios have gradually become familiar to people: "A freight driver no longer seeks customers offline, but installs an application on his mobile phone, and receives orders through the freight platform, which can reduce the unloaded ratio of his truck and ensure the payment of transportation expenses, which not only reduces costs but also increases income." Behind the freight platform is the exchange of information between hundreds of thousands of freight drivers and shippers every day, the transportation value of over 100 million, the track coordinates of millions of trucks, hundreds of millions of driving data, and PB-level data volume. "In the post-COVID-19 epidemic era, online learning has become a normal state. A primary school student takes online classes online, communicates with teachers without delay, and uses various teaching aids to learn knowledge." The online education system providers need to face millions of daily active users every day, provide more than 100 kinds of teaching functions and teaching aids, covering six continents around the world, and provide 1080P high-definition live broadcast, and the delay needs to be less than 200ms. "On June 18 of the year, JD's merchants are making great efforts to actively prepare for sales promotion. A merchant is updating