News for hackers. Stuff that matters.
Re:Time Structured Merge Tree: From LSM Tree to B+Tree and B... by ciconia on Wednesday October 07, 2015 @03:01PM (#10348196)
Try running that benchmark again on a table that already contains 500M records. I'm quite sure you'll see different numbers. Another problem you'll run into is reading back those records, and performing aggregations and window functions on your data. In my experience Postgresql is not a bad solution for time series data, but it's certainly not a perfect fit.
Re:Time Structured Merge Tree: From LSM Tree to B+Tree and B... by dkhenry on Wednesday October 07, 2015 @03:40PM (#10348480)
If I take the index off the benchmark will run just about as fast for 500M rows as it did for 0 rows, at 500M rows MySQL's BTREE index starts to fall over, but I have plenty of options to mitigate that. I don't think anything is a perfect fit for time series data, but again my point was that the traditional relational database's are _really_ good, and they do in fact scale up well beyond what most people think they will.