OPENTSDB Introduction--a distributed, scalable time-series database based on HBase, while HBase essence is Columnstore

original link: http://www.jianshu.com/p/0bafd0168647OPENTSDB Introduction 1.1, Opentsdb is what? What are the main uses?

Official documentation describes this: Opentsdb is a distributed, scalable time Series Database (TSDB) written on top of HBase;

A distributed, scalable time-series database based on HBase is translated.

The main purpose is to do a monitoring system, such as the collection of large-scale clusters (including network equipment, operating systems, applications) monitoring data and storage, query.

1.2. Introduction Continue

The data stored in the OPENTSDB is in metric, metric is a monitoring item, such as the server, there will be CPU utilization, memory utilization of these metric;

OPENTSDB uses hbase as storage, and because of its good design, the data storage for metric is supported to the second level;

OPENTSDB supports persistent data storage, that is, the saved data is not deleted voluntarily, and the original data is saved (some monitoring systems will be saved after the data has been aggregated longer)

OPENTSDB Storage-related concepts

When we introduce these concepts, we will look at a practical scenario first.

For example, suppose we collect 1 servers (hostname=qatest) CPU utilization, found that the server at 21:00, the CPU utilization rate reached 99%

Take a look at some of the core concepts of OPENTSDB storage with examples

1) Metric: That is what we call the monitoring items. For example, the CPU utilization above

2 Tags: is some tags, in opentsdb inside, tags by tagk and tagv composition, that is tagk=takv. Tags are used to describe metric, such as the above for the mark is server a cpuusage,tags can be hostname=qatest

3) Value: A value that represents the actual value of a metric, such as the 99% above

4) Timestamp: A timestamp that describes when value is, for example, 21:00

5) Data point: The value of a metric at a certain time.

The Data point includes the following sections: Metric, Tags, Value, Timestamp

The server described above at 21:00 CPU utilization, is a datapoint

Saved to the Opentsdb, is countless datapoint.

Opentsdb's overall architecture
OPENTSDB Architecture

Servers: Is the server, the above C refers to collector, can be understood as OPENTSDB agent, through the collector collect data, push data;

TSD:TSD is a stateless server for external communication, collector can push the monitoring data through the simple RPC protocol of TSD, and TSD also provides a Web UI page for data query, as well as a script to query the monitoring data, alarm monitoring data

HBASE:TSD received monitoring data, it is through asynchbase this library to write data to Hbase;asynchbase is a fully asynchronous, non-blocking, thread-safe hbase client, with fewer threads, locks and memory, can provide higher throughput, Especially for a lot of write operations.

OPENTSDB Introduction--a distributed, scalable time-series database based on HBase, while HBase essence is Columnstore