Re:Time-Series Database Requirements
by Rapzid on Sunday March 08, 2015 @11:22PM
(#9168827)
I would actually be very interested in hearing more about this MySQL implementation.
[ Parent
| on HN ]
I work with Baron, the author. He gave a talk recently at SCaLE[0] about our time series storage using MySQL. The slides[1] are also available in case you don't want to watch the entire talk.
[0] http://youtu.be/EoUfkkrIbPg
[1] http://www.slideshare.net/vividcortex/scaling-vividortexs-bi...
Thanks very much. I have experimented with used Riak as a backend with its search 2.0(lucene) for indexing the streams. I'm very interested in how the MySQL tables are laid out; On the watch/reading list for tonight :)
I'm wondering if it has been considered to buffer the metrics into a memory table so they can participate in queries, and then let MySQL flush them to durable tables at regular intervals?
My personal opinion on that is that it's another layer of complexity without too much to gain. It's something else to manage (operational overhead), and something else to interact with (programmer overhead).
MySQL by itself is plenty fast for us. As Baron mentioned, we don't even run the database servers at full capacity.