The other big factor is that almost all open source database engines are the product of someone who is basically designing their first database engine. Prometheus is now on its third(?) redesign, and while you can see the growth of the designers' skill it is still a relatively naive design. This is not a jab at the designers, it just takes many creative iterations over many years to discover all the tricks that the experts know but nobody publishes. I probably spent a decade producing database engines that in hindsight were pretty mediocre but at the time I thought they were sophisticated -- I didn't know what I didn't know.
As a last observation, the other issue is that building a genuinely sophisticated database engine requires a commitment to writing an enormous amount of code before the barest database kernel can even bootstrap, on the order of 20kLoC of dense C++, never mind provide any database functionality. I've noticed that most open source projects need to see the payoff of code running and doing something minimally useful with much lower levels of investment. For database engines, getting to running code as quickly as possible compromises the design but the pressure to get to running code is understandable. The amount of time and effort required to design and build a state-of-the-art database engine goes far beyond what most people are willing to invest in what is essentially a hobby.
This has been true of network programming for a long time as well. Private companies find novel ways of switching and routing packets, reducing latency, etc. and academia is left to pick up the pieces.
This makes an interesting case for software patents: what if database companies patented those insights instead of keeping them secret, and thus could keep others from using them for, say, 3-4 years? At the end of that time anyone could use those techniques.
This seems to me better than either trade secrets are the current absurdly-long patent lifetimes.
Typically these solutions tend to follow the money, but if you ignore the list price, you can usually get a decent deal for a startup.
I'm only slightly familiar with Quasar, but I believe it is just a key-value store with document characterists, not a general database. The comparisons on their website are mostly to other KV stores and document systems.
You can have a look what happend with extremely simple code bases over time. Imagine the same effect on larger code.
https://www.computer.org/cms/Computer.org/ComputingNow/issue...
However I am interested to know why this kind of data and/or problems require specific software and why it can't be handled by traditional RDBMs? Obviously you could model the domain with a classic database, but seemingly there exist important queries that can't be satisfied (at least timely) through classic systems - what kind of queries are these, and why do they fail using a general-purpose DB approach?
Most Time-series databases are columnar in nature, and often have the concept of time baked into the ordering of values (think vectors not sets). Because they are columnar, they are more trivial to retrieve just the data needed by a query. Suddenly instead of loading a billion rows and averaging the value in one column, you're just accesing the column itself to answer the question. From an IO perspective that's a huge savings.
Now imagine you have special, dark arts for working on compressed data, and a query optimizer you've been tuning for a decade for demanding clients. It does not surprise me that kdb is much faster than the open source competitors. And to be fair, even with excellent traditional databases like postgres, I bet Oracle, db2, and Ms SQL are still generally faster in most queries.
The other side of this is that writing out data in that form would naively be an iop per sample, as you're usually appending one sample to 100s of time series in one request.
A significant part of monitoring TSDB design is buffering up samples and batching writes in order reduce that iop rate to something sane.
For example in the right circumstances the 1.x Prometheus design can ingest 250k samples/s on a hard disk which provides ~100 iops/s.
https://blog.timescale.com/when-boring-is-awesome-building-a...
There's Tgres, that stores time series in Postgres, efficiently (using Postgres arrays and views):
https://github.com/tgres/tgres
https://grisha.org/blog/2017/01/21/storing-time-seris-in-pos...
I've actually never heard of ExtremeDB. It's only in-memory though, so it can it even handle the huge datasets that timeseries/column dbs handle? I don't see any performance tests for it either.
http://financial.mcobject.com/downloads/white-papers-collate...
Why should we give that work away for free, especially if it's not that hard to roll your own for most small-scale needs?
Then maybe they should learn how to write their own databases, or pay people for their expertise. :)
Is it that people want to make the problem more complex that it needs to be? Is it that those who know most about these issues don't share their secrets so implemented from the outside often don't have a good understanding of how to do things properly? If you were to asked the guy behind Prometheus if he's looked at the commercial offerings and what he's learned from them, would even be able to speak about them intelligently?
There seems to be a huge skills gap on these things that I can't put my finger on. I'd love to be able to use a real TSDB, even at only half the speed and usefulness. It would be great for these smaller firms that cant or wont pay the license fees for a commercial offering until they get larger.