Solution highlights: This solution can handle petabytes of data and analyze both online data and offline data.
Behavioral logs collected from game servers must be stored and analyzed to calculate statistics, such as the player retention rate, life time value (LTV), average revenue per user (ARPU), and the total recharge amount.
Game operations plans change frequently. Therefore, a flexible schema is required.
Large numbers of online users generate a large amount of log data. Such log data requires a high-throughput and cost-effective processing platform.
Game masters (GMs) need to check the raw data and may also need to analyze data immediately after the data is generated. The business requirements for both online and offline business scenarios must be met.
ApsaraDB for Lindorm (Lindorm) can adapt to rapid business changes. It supports flexible schema and dynamic columns.
Lindorm can write volumetric data in real time, concurrently process millions of or tens of millions of requests per second, and store petabytes of data. Lindorm stores hot and cold data into different storage media. Log data is directly written to and read from Object Storage Service (OSS) buckets. This reduces the log storage costs.
Lindorm is seamlessly integrated with the Spark engine of Data Lake Analytics (DLA) to support data processing, job scheduling, SQL execution, and building offline data warehouses. DLA Spark Streaming is used to extract, transform, and load (ETL) data in real time and write the results into Lindorm. The DLA Spark engine is used to analyze the reports and other data of the game platform offline.
After the data is migrated to Lindorm, Lindorm efficiently manages the rapid business growth and also provides highly concurrent write capability. This facilitates the operational decision-making of the company.
Lindorm is integrated with the DLA Spark engine to meet the requirements of the game platform, such as real-time ETL and offline analysis. This way, a closed-loop system is formed to manage various requirements.