3 ms·

Database in Go is 6x faster than the database in Rust

6 points by valyala 6y ago

ibraheemdev 6y ago
You cannot compare two languages by benchmarking two databases that happen to be written in those languages. You are comparing victoriametrics, and influx IO. You are not comparing Go and Rust.
HereBeBeasties 6y ago
More to the point you are comparing something that's only just been released and probably hasn't been tuned yet against something that has. The title on this post is extremely misleading. It should read "mature timeseries storage code outperforms immature timeseries storage code". But then, I guess no one would click on that?
bionhoward 6y ago
All these database companies focus way too much on benchmarks and not enough on simple stuff like docs, autoscaling / serverless pricing models, examples on git, permissive license. Redis is a sad example. Nice open source library but the Redis Labs stuff is pretty much a joke, because to get the modules like graphs, JSON, search, it costs thousands of dollars. And the “source available” license is stupid, “you can’t use this” - the Open Source equivalent of “look but don’t touch” ... not to mention they have so many different pricing levels it becomes a massive decision which one to use (high friction purchase) ... but hey, it’s fast, right? FaunaDB is much more compelling than Redis Enterprise because of serverless pricing, but uses some custom query language, yikes. Likewise DynamoDB has nice pricing but way complicated / proprietary interfaces. The Cassandra service could be cool. Nice benchmarks will never make up for craptastic / complex pricing models, lame license, bad docs / examples. Too many nerds think if they post higher numbers, it makes up for massive failure at simple business. Need more nerds who keep it simple, documented, autoscaling, easy to sign up, etc The fact there is no great serverless graph db and so few viable auth services in 2020 is a shameful travesty for the industry... </rant>

6 points by valyala 6y ago

You cannot compare two languages by benchmarking two databases that happen to be written in those languages. You are comparing victoriametrics, and influx IO. You are not comparing Go and Rust.

HereBeBeasties 6y ago

More to the point you are comparing something that's only just been released and probably hasn't been tuned yet against something that has. The title on this post is extremely misleading. It should read "mature timeseries storage code outperforms immature timeseries storage code". But then, I guess no one would click on that?

bionhoward 6y ago

All these database companies focus way too much on benchmarks and not enough on simple stuff like docs, autoscaling / serverless pricing models, examples on git, permissive license. Redis is a sad example. Nice open source library but the Redis Labs stuff is pretty much a joke, because to get the modules like graphs, JSON, search, it costs thousands of dollars. And the “source available” license is stupid, “you can’t use this” - the Open Source equivalent of “look but don’t touch” ... not to mention they have so many different pricing levels it becomes a massive decision which one to use (high friction purchase) ... but hey, it’s fast, right? FaunaDB is much more compelling than Redis Enterprise because of serverless pricing, but uses some custom query language, yikes. Likewise DynamoDB has nice pricing but way complicated / proprietary interfaces. The Cassandra service could be cool. Nice benchmarks will never make up for craptastic / complex pricing models, lame license, bad docs / examples. Too many nerds think if they post higher numbers, it makes up for massive failure at simple business. Need more nerds who keep it simple, documented, autoscaling, easy to sign up, etc The fact there is no great serverless graph db and so few viable auth services in 2020 is a shameful travesty for the industry... </rant>