osrg / bgperf Goto Github PK

View Code? Open in Web Editor NEW

86.0 14.0 30.0 348 KB

License: Apache License 2.0

Python 100.00%

bgperf's Introduction

bgperf

bgperf is a performance measurement tool for BGP implementation.

Prerequisites

Python 2.7 or later
Docker

How to install

$ git clone https://github.com/osrg/bgperf
$ cd bgperf
$ pip install -r pip-requirements.txt
$ ./bgperf.py --help
usage: bgperf.py [-h] [-b BENCH_NAME] [-d DIR]
                 {doctor,prepare,update,bench,config} ...

BGP performance measuring tool

positional arguments:
  {doctor,prepare,update,bench,config}
    doctor              check env
    prepare             prepare env
    update              pull bgp docker images
    bench               run benchmarks
    config              generate config

optional arguments:
  -h, --help            show this help message and exit
  -b BENCH_NAME, --bench-name BENCH_NAME
  -d DIR, --dir DIR
$ ./bgperf.py prepare
$ ./bgperf.py doctor
docker version ... ok (1.9.1)
bgperf image ... ok
gobgp image ... ok
bird image ... ok
quagga image ... ok

How to use

Use bench command to start benchmark test. By default, bgperf benchmarks GoBGP. bgperf boots 100 BGP test peers each advertises 100 routes to GoBGP.

$ sudo ./bgperf.py bench
run tester
tester booting.. (100/100)
run gobgp
elapsed: 16sec, cpu: 0.20%, mem: 580.90MB
elapsed time: 11sec

To change a target implementation, use -t option. Currently, bgperf supports BIRD and Quagga other than GoBGP.

$ sudo ./bgperf.py bench -t bird
run tester
tester booting.. (100/100)
run bird
elapsed: 16sec, cpu: 0.00%, mem: 147.55MB
elapsed time: 11sec
$ sudo ./bgperf.py bench -t quagga
run tester
tester booting.. (100/100)
run quagga
elapsed: 33sec, cpu: 0.02%, mem: 477.93MB
elapsed time: 28sec

To change a load, use following options.

-n : the number of BGP test peer (default 100)
-p : the number of prefix each peer advertise (default 100)
-a : the number of as-path filter (default 0)
-e : the number of prefix-list filter (default 0)
-c : the number of community-list filter (default 0)
-x : the number of ext-community-list filter (default 0)

$ sudo ./bgperf.py bench -n 200 -p 50
run tester
tester booting.. (200/200)
run gobgp
elapsed: 23sec, cpu: 0.02%, mem: 1.26GB
elapsed time: 18sec

For a comprehensive list of options, run sudo ./bgperf.py bench --help.

bgperf's People

Contributors

Stargazers

Watchers

bgperf's Issues

Could not run on FreeBSD

Hello, folks!

I would like to use toolkit on FreeBSD but I could not start it:

./bgperf.py --help
Traceback (most recent call last):
  File "./bgperf.py", line 30, in <module>
    from nsenter import Namespace
  File "/usr/local/lib/python2.7/site-packages/nsenter/__init__.py", line 21, in <module>
    class Namespace(object):
  File "/usr/local/lib/python2.7/site-packages/nsenter/__init__.py", line 52, in Namespace
    _libc = ctypes.CDLL('libc.so.6', use_errno=True)
  File "/usr/local/lib/python2.7/ctypes/__init__.py", line 365, in __init__
    self._handle = _dlopen(self._name, mode)
OSError: Shared object "libc.so.6" not found, required by "python"

Unfortunately, we haven't nsenter and cgroups at FreeBSD. Could you offer workaround for this case?

You're benching Quagga with a 30s MRAI against implementations with no MRAI

Hi,

You're bench marking Quagga in its default configuration with a 30s MRAI against BGP implementations that do not implement an MRAI at all. Your convergence time numbers as a result are not apples-apples.

You should configure Quagga with a 0s MRAI (neighbor ... advertisement-interval), or at least some part of your numbers are just indicative of what no-MRAI BGP versus 30s MRAI.

Python exception thrown when running bench

I've just installed bgperf, as described in your instructions.

21:18:49 [dotwaffle@hasselhoff:~/Code/bgperf] master ± sudo ./bgperf.py bench
creating Docker network "bgperf-br" with subnet 10.10.0.0/16
run monitor
run gobgp
waiting bgp connection between gobgp and monitor
run tester tester type normal
tester booting.. (100/100)
elapsed: 0sec, cpu: 0.00%, mem: 0.00B, recved: 9900
Exception in thread Thread-2:
Traceback (most recent call last):
  File "/usr/lib/python2.7/threading.py", line 810, in __bootstrap_inner
    self.run()
  File "/usr/lib/python2.7/threading.py", line 763, in run
    self.__target(*self.__args, **self.__kwargs)
  File "/srv/home/dotwaffle/Code/bgperf/base.py", line 179, in stats
    queue.put({'who': self.name, 'cpu': cpu_percentage, 'mem': stat['memory_stats']['usage']})
KeyError: 'usage'
elapsed: 2sec, cpu: 0.00%, mem: 0.00B, recved: 10000
21:20:20 [dotwaffle@hasselhoff:~/Code/bgperf] master ±

Any ideas? I made sure to run pip --upgrade -r pip-requirements.txt in case that helps... It didn't. Also happens using -t bird.

System is Debian Stable.

Be nice to support HTTP proxies

It'd be nice if the docker build of the containers supported passing in a HTTP proxy somehow, for the apt commands, for those behind a corp firewall.

For now I just manually added in ENV http_proxy ... etc to the {bird,exabgp,gobgp,quagga}.py files.

regards,

Paul

osrg / bgperf Goto Github PK

bgperf's Introduction

bgperf

Prerequisites

How to install

How to use

bgperf's People

Contributors

Stargazers

Watchers

Forkers

bgperf's Issues

Could not run on FreeBSD

You're benching Quagga with a 30s MRAI against implementations with no MRAI

Python exception thrown when running bench

Be nice to support HTTP proxies

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent