tencent cloud

Apache Druid is a distributed data processing system supporting real-time and multi-dimensional online analytical processing (OLAP). It is used to implement quick and interactive query and analysis for large data sets.

Basic Characteristics

Characteristics of Apache Druid:

• It supports interactive queries with a subsecond response time and has various features such as multi-dimensional filtering, ad hoc attribute grouping, and quick data aggregation.
• It supports highly concurrent and real-time data ingestion to ensure real-timeliness for data ingestion and query.
• It features high scalability. With the distributed shared-nothing architecture, it supports quick processing of petabytes of data with hundreds of billions of events and sustains thousands of concurrent queries per second.
• It allows simultaneous online queries by multiple tenants.
• It supports high availability (HA) and rolling update.

Use Cases

Druid is most frequently used for flexible, quick, multi-dimensional OLAP analysis on big data. In addition, as it supports ingestion of pre-aggregated data and analysis of aggregated data based on timestamps, it is usually used in time-series data processing and analysis, such as ad platform, real-time metric monitoring, recommendation model, and search model.

System and Architecture

Druid uses a microservice-based architecture. All core services in it can be deployed on different hardware devices either separately or jointly.

Enhanced EMR Druid
A lot of improvements have been made on EMR Druid based on Apache Druid, including integration with EMR Hadoop and relevant Tencent Cloud ecosystem, convenient monitoring and OPS, and easy-to-use product APIs, so that you can use it out of the box in an OPS-free manner.

Currently, EMR Druid supports the following features:

• Easy integration with EMR Hadoop cluster
• Easy and quick elastic scalability
• HA
• Using COS as deep storage
• Using COS file as data source for batch indexing
• Metadata storage in TencentDB
• Integration with tools such as Superset
• Various monitoring metrics and alarm rules
• Failover
• High security