time series database, prometheus opentsdb influxdb Graphite

db.md

CAAS platform APP resource monitoring

APP resource monitoring

• Data should be kept as long as Long time apps running
• Data should be kept months after apps die
• Support querying data per container
• Support querying data per app
• Support querying data per cluster
• Tens of thousands of collectors
• Load balancing multiple query engines

Available DB storage

comparision between prometheus opentsdb influxdb Graphite

name	design	performance	scale	availability	usage
OpenTSDB	Hbase	A single TSD can handle thousands of writes per second, millions of writes per second	tens of thousands	HBase HA + Haproxy	http://www.searchtb.com/2012/07/opentsdb-monitoring-system.html
Influxdb	rocksdb/leveldb/SQL	-	pre-configed shards	Haproxy + double write	-
Prometheus	Private storage	-	require explicit sharding once the capacity of a single node is exceeded	-	-
Graphite		-	-	-	-
Zabbix	RDBMS mysql/pgsql	-	-	-	-

http://db-engines.com/en/ranking/time+series+dbms

OpenTSDB

Tag方式存储数据 start=1356998400&m=sum:sys.cpu.user{host=webserver01} sys.cpu.user host=webserver01,cpu=0

8 tags at most

3 bytes uid

Available Collectors

cadvisor

Reference

http://db-engines.com/en/ranking

https://influxdata.com/blog/benchmarking-leveldb-vs-rocksdb-vs-hyperleveldb-vs-lmdb-performance-for-influxdb/