Zabbix is a great tool and i use it since 2009, but i think there are some things that can we much better 
Soo, zabbix supported a lot of databases, but why? Its more complex to develop for different databases, i think MariaDB/Mysql, Postgres and SQlite (only for proxys) are enough,
so you can tune the querys and use some database specific features. For example hstore and JSON in Postgres to increase the performance, also you can use PostgreSQL to use ltree maybe fo subgroups..
or use serial/auto_increment instead of IDs (ZBXNEXT-1006).
So far i know Zabbix SIA think about to use a NoSQL Database for history and trends, but why, Postgres has also nosql features (hstore, JSON(B) ) or MariaDB has dynamic columns or
you can the storage-engine cassandra.
So the most performance-killer are big databases, but it exist some ideas du decrease values and increase the performance, for example: ZBXNEXT-2436 or ZBXNEXT-1846 but both are closed by a duplicate to each other?!
Furthermore it exist some table-partitons mechanism, why dont support zabbix this?
I think zabbix must drop some database-support and must require newer version to support alle this features.. use mysql >5.5 (it exist since 2010), and also PostgreSQL >9.1, older versions are mostly EOL.
Soo, zabbix supported a lot of databases, but why? Its more complex to develop for different databases, i think MariaDB/Mysql, Postgres and SQlite (only for proxys) are enough,
so you can tune the querys and use some database specific features. For example hstore and JSON in Postgres to increase the performance, also you can use PostgreSQL to use ltree maybe fo subgroups..
or use serial/auto_increment instead of IDs (ZBXNEXT-1006).
So far i know Zabbix SIA think about to use a NoSQL Database for history and trends, but why, Postgres has also nosql features (hstore, JSON(B) ) or MariaDB has dynamic columns or
you can the storage-engine cassandra.
So the most performance-killer are big databases, but it exist some ideas du decrease values and increase the performance, for example: ZBXNEXT-2436 or ZBXNEXT-1846 but both are closed by a duplicate to each other?!
Furthermore it exist some table-partitons mechanism, why dont support zabbix this?
I think zabbix must drop some database-support and must require newer version to support alle this features.. use mysql >5.5 (it exist since 2010), and also PostgreSQL >9.1, older versions are mostly EOL.
