Baidu MapReduce’s hybrid deployment solution allows for “time-division” use of BCC. In other words, BCC focuses on its own business when it is busy and helps the BMR cluster’s big data computing when it is idle. This solution improves BCC’s resources usage and enhances the BMR cluster’s computing ability, generally reducing IT expenditures.

A famous Internet enterprise in China rents BCC to build business systems and creates a BMR cluster to massively compute daily business data, as the following figure shows:

Wave peak and wave trough are distinct to visits of business systems such as e-commerce and email. To ensure normal business operations, the enterprise purchases BCC resources based on maximum business needs. The website visits are quite low in the early morning, and BCC resources to support business systems are much wasted.

BMR hybrid deployment solution can solve it. As the following figure shows, when resource usage by business systems is low in the early morning, the idle BCC computing resources are assigned to BMR big data computing to enhance the BMR cluster’s computing ability; when the visits increase, the BCC computing resources are separated from BMR system. Through peak-load shifting, the time-division use of BCC resources ensures maximum utilization of resources and dramatically reduces total costs.

Essentially, the BMR hybrid deployment solution uses BCC to expand the capacity of the BMR cluster. After capacity expansion, the cluster includes BMR nodes and BCC nodes, and BCC nodes are equal to task nodes of the BMR cluster and enhance the cluster’s computing ability. After capacity expansion, the Master node equally treats task nodes from BCC or BMR when assigning tasks, and BMR provides the same step-level fault tolerance to task nodes.

For further understanding of the BMR hybrid deployment solution, please contact bce@baidu.com or submit Ticket, and Baidu AI Cloud offers complete solutions to you.