mirror of https://github.com/SpeckiJ/Hochwasser
You cannot select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
75 lines
2.7 KiB
Markdown
75 lines
2.7 KiB
Markdown
<h1 align="center" >🌊🌊🌊 Hochwasser 🌊🤽🌊</h1>
|
|
<p align="center"><img src="benchmarks/hochwasser_shuffle_vs_ordered.gif"/></p>
|
|
<p align="center"><img src="benchmarks/nmzs.gif"/></p>
|
|
|
|
Highly efficient distributed [Pixelflut] client.
|
|
|
|
- Sends static images, text, generated patterns (animations upcoming)
|
|
- REPL enables fast iterations
|
|
- CnC server + client architecture (it's webscale!) (can also run in a single process)
|
|
- Faster than [sturmflut] (in some benchmarks at least)
|
|
- No dependencies (pixelflut apparently was considered a primary use case in the design of golang's stdlib 👍)
|
|
|
|
[pixelflut]: https://cccgoe.de/wiki/Pixelflut
|
|
[sturmflut]: https://github.com/TobleMiner/sturmflut
|
|
|
|
### testimonials
|
|
> Pixelflut endlich *durchgespielt*.
|
|
>
|
|
> - Steffen Cybert
|
|
|
|
> N-no more micro-ddosing: bring on ssome Hochwasser and exppperience colors never seen befffore!1!
|
|
>
|
|
> - Morty
|
|
|
|
> Hochwasser brings back the D in *social DDoSing*! Man, I forgot which one..
|
|
>
|
|
> - Doc Brown
|
|
|
|
## build / install
|
|
1. have a `go` installation >= 1.12
|
|
2. `go get github.com/SpeckiJ/Hochwasser`
|
|
3. `go install github.com/SpeckiJ/Hochwasser`
|
|
|
|
> The help texts may be lacking, it's recommended to read `rpc/repl.go`.
|
|
|
|
## hacking
|
|
Look at the `github.com/SpeckiJ/Hochwasser/pixelflut` subpackage, it contains the performance sensitive core.
|
|
|
|
The code is getting somewhat ~~bloated~~enterprise-ready, so if you want to quickly render a fun thing,
|
|
it may be easier to just build a separate executable on top of `pixelflut.Flut()`, than to extend Hochwasser.
|
|
|
|
## benchmark
|
|
The following benchmark was run on a max-spec X280 against version [d4c574b].
|
|
|
|
I could not figure out what the performance bottleneck is, but it doesn't seem
|
|
to be CPU limited, as turbo-boost doesn't kick in.
|
|
|
|
To reproduce, run the following commands in separate shells:
|
|
|
|
```sh
|
|
iperf -s -p 1234
|
|
go run main.go -image benchmark/test.png -connections 10
|
|
```
|
|
|
|
|
|
|
|
55 Gbps on average! 🌊🌊🌊
|
|
|
|
[sturmflut] (`./sturmflut 127.0.0.1:1337 benchmark/test.png -t 10`, version `8ec6ee9`) managed to get 48 Gpbs throughput on this system.
|
|
|
|
> Hint: Benchmarking throughput against the [pixelnuke][pixelflut_gh] server is
|
|
pointless, as performance is then CPU-limited to ~1 Gbps by the server.
|
|
Using [iperf] removes the server limitation.
|
|
This also means that these metrics of several Gbps are far higher than
|
|
realworld scenarios.
|
|
|
|
[d4c574b]: https://github.com/SpeckiJ/Hochwasser/commit/d4c574be103a7bad69349f29402694f51058184c
|
|
[pixelflut_gh]: https://github.com/defnull/pixelflut
|
|
[iperf]: https://iperf.fr/
|
|
|
|
## future ideas
|
|
see [IDEAS](https://github.com/SpeckiJ/Hochwasser/blob/master/IDEAS.md).
|
|
|
|
<p align="center"><img src="benchmarks/hochwasser_vs_sturmflut.gif"/></p>
|