Druid是一个为流式数据设计的实时分析数据存储系统，包含4个组件，采用了冷热分离的结构：
Real-time Nodes
Historical Nodes
Broker Nodes
Coordinator Nodes
infoq里有一篇介绍druid的文章。
Druid is similiar to C-Store [38] and LazyBase [8] in that it has
two subsystems, a read-optimized subsystem in the historical nodes
and a write-optimized subsystem in real-time nodes. Real-time nodes
are designed to ingest a high volume of append heavy data, and do
not support data updates. Unlike the two aforementioned systems,
Druid is meant for OLAP transactions and not OLTP transactions.
Druid’s low latency data ingestion features share some similarities
with Trident/Storm [27] and Spark Streaming [45], however,
both systems are focused on stream processing whereas Druid is
focused on ingestion and aggregation. Stream processors are great
complements to Druid as a means of pre-processing the data before
the data enters Druid.
There are a class of systems that specialize in queries on top of
cluster computing frameworks. Shark [13] is such a system for
queries on top of Spark, and Cloudera’s Impala [9] is another system
focused on optimizing query performance on top of HDFS. Druid
historical nodes download data locally and only work with native
Druid indexes. We believe this setup allows for faster query latencies.
Druid leverages a unique combination of algorithms in its architecture.
Although we believe no other data store has the same set
of functionality as Druid, some of Druid’s optimization techniques
such as using inverted indices to perform fast filters are also used in
other data stores [26].
