chaotic-good

10 ms·

31.
by chaotic-good 9y ago
Prometheus is a monitoring system (pull based), Akumuli is a TSDB (push based). I believe that one can use Akumuli as a long-term storage for Prometheus.
32.
by chaotic-good 9y ago
Akumuli is quite different from InfluxDB. It focuses on single node performance and operational simplicity. Essentially I'm trying to make it a "fire and forget" kind of app. No idea about Druid.
33.
Storage engine design(akumuli.org)
179 points by chaotic-good 9y ago|28 comments
34.
Running time-series database on 32-core machine (16M writes/second)(akumuli.org)
1 points by chaotic-good 10y ago|0 comments
35.
Benchmarking Akumuli – 4.5M writes per second on AWS(akumuli.org)
4 points by chaotic-good 10y ago|0 comments
36.
by chaotic-good 10y ago
I saw this first time many years ago. It clearly states that exceptions is a good thing but shouldn't be used in Google's projects because their code is not prepared for exceptions: > On their face, the benefits of using except
37.
by chaotic-good 10y ago
This article doesn't have a clue. For example: > but many C++ programmers disable exceptions for performance I have never seen a project like this for 10+something years. The only piece of code with disabled exceptions that I'v
38.
Predictive compression algorithm for time-series(akumuli.org)
2 points by chaotic-good 10y ago|0 comments
39.
by chaotic-good 10y ago
Sequitur
40.
Akumuli Markedly Outperforms InfluxDB in Time-Series Data and Metrics Benchmark(akumuli.org)
11 points by chaotic-good 10y ago|0 comments
41.
by chaotic-good 10y ago
> Exceptions give you stack traces automatically. All of that context (and more) is there without library authors having to manually weave it in at every level of calls. I find exceptions really useful but also, I think that with excepti
42.
by chaotic-good 10y ago
Many infrastructure services don't have any auth at all. This doesn't make them bad. This just means that these products have been developed for trusted environments. Even if MongoDB would have been configured properly by default
43.
by chaotic-good 10y ago
Error checks are almost free and can be considered zero cost because branch predictor works really well with error checks (because errors are rare). Branch predictor uses some space in L1 cache to store tables so it's not 100% free but
44.
by chaotic-good 10y ago
>> No(Auth)SQL. Absence or presence of auth is irrelevant. You're database servers, message queues and other infrastructure shouldn't be accessible from the internet. No auth protocol can protect you from this.
45.
Time-series compression (part 1)(akumuli.org)
3 points by chaotic-good 10y ago|1 comments
46.
by chaotic-good 10y ago
>> In most genres people would rather have the simulation pause when there is packet loss, and then run fast to catch up rather than have unreliable data about player locations. I thought that modern games use interpolation to compens
47.
The UCR Matrix Profile Page(cs.ucr.edu)
2 points by chaotic-good 10y ago|0 comments
48.
by chaotic-good 10y ago
I'm working on my own solution for a while (about 3 years) - https://github.com/akumuli/Akumuli . One of the greatest things about it is that there is no need for preconfigured rollups. The storage is based on B+Tr
49.
by chaotic-good 10y ago
https://en.wikipedia.org/wiki/Unevenly_spaced_time_series You can call that _logging_ if you want, but people will ask for it anyway.
50.
by chaotic-good 10y ago
> I do not understand the general obsession with specialized TS stores Specialized TS stores exist for decades (process historians, tickerplants, and so on).
51.
by chaotic-good 10y ago
Calling malloc is not that expensive if you're using jemalloc or any other modern allocator. And with generational GC allocation is not just a register increment because Java has mandatory object initialization and all objects will be
52.
by chaotic-good 10y ago
I've found several submitted issues that described this problem. All this issues are closed now so maybe I was wrong about Bolt.
53.
by chaotic-good 10y ago
BoltDB can be corrupted on crash (kill -9 or hard reset) at least this was the case when I gave it a try. Any database can but in that case that was easy to achieve.
54.
by chaotic-good 10y ago
You can try LMBD - https://symas.com/products/lightning-memory-mapped-database/ In my opinion there is exactly two production ready embedded databases - SQLite and LMDB. At least this two databases have decent cra
55.
by chaotic-good 10y ago
I'm glad that I was wrong about it.
56.
by chaotic-good 10y ago
It is usually a good idea to discard late writes. It should be possible to write data with some time-stamp skew but if skew is too large data should be discarded as erroneous. It's totally depends on an application so.
57.
by chaotic-good 10y ago
I think that compression is a must for the serialization library. Protobuf uses almost twice less memory than Cap'n Proto. Using an external compression is not an option in some cases. E.g. consider building tcp-server that communicate
58.
by chaotic-good 10y ago
With C++ one can go far further than that. It's the matter of time spent on the solution. With all that other languages there is a performance ceiling that can't be breached.
59.
by chaotic-good 10y ago
I'm working on my own time-series database server here - https://github.com/akumuli/Akumuli The most limiting factor is not the query language or network its a storage engine itself (and not the HDD/SSD writ
60.
by chaotic-good 10y ago
I find QtCreator superior to CLion. It's faster and more polished, has better vi-emulation and clang ecosystem integration.

More ›