DEV Community

Discussion on: Choosing TSDBs - InfluxDB for Us

Collapse
manigandham profile image
Mani Gandham

Might as well just use something like MemSQL and get a real distributed relational database with columnstore functionality. Works great with time-series or any other data warehouse needs with full SQL support, including joins (and fast in-memory rowstore tables for OLTP).

Collapse
stevemushero profile image
Steve Mushero Author

Yes, InfluxDB certainly needs an HA / distributed solution, but using non-TSDB like this means losing tagging, protocol support (UDP rules!), alerting options, data aggregation, dimensional / time analyses, and probably more like high cardinality, multi-tag indexing on large data sets - especially with some newer TSDBs really diving into stats functions for clustering, anomalies, etc. MemSQL looks interesting, a bit complex, MySQL wire-compat is nice. I'm pretty new to TSDBs, too, so probably missing stuff.

Collapse
sdracerunner profile image
Tony

Influx does have a HA solution, but it's closed source and had associated license costs. We use it in our setup with a load balancer to front the HTTP endpoint and it works pretty well. Search for Influx Enterprise for details, all the docs are public.

Thread Thread
stevemushero profile image
Steve Mushero Author

Yes, they do have a new commercial HA system, though seems quite young and I think was created after another attempt, so still a ways to go, plus of course would prefer to see an open-source edition, even if had limited shards, size, etc. so lots of people could run it for real in real systems.