how_bgperf_works.md

How bgperf works

When bench command issued, bgperf boots three (or more) docker containers, target, monitor and one or more tester and connect them via a bridge (bgperf-br by default).

By default, bgperf stores all configuration files and log files under /tmp/bgperf. Here is what you can see after issuing bgperf.py bench -n 10.

$ tree /tmp/bgperf
/tmp/bgperf
├── gobgp
│   ├── gobgpd.conf
│   ├── gobgpd.log
│   └── start.sh
├── monitor
│   ├── gobgpd.conf
│   ├── gobgpd.log
│   └── start.sh
├── scenario.yaml
└── tester
    ├── 10.10.0.10.conf
    ├── 10.10.0.10.log
    ├── 10.10.0.11.conf
    ├── 10.10.0.11.log
    ├── 10.10.0.12.conf
    ├── 10.10.0.12.log
    ├── 10.10.0.3.conf
    ├── 10.10.0.3.log
    ├── 10.10.0.4.conf
    ├── 10.10.0.4.log
    ├── 10.10.0.5.conf
    ├── 10.10.0.5.log
    ├── 10.10.0.6.conf
    ├── 10.10.0.6.log
    ├── 10.10.0.7.conf
    ├── 10.10.0.7.log
    ├── 10.10.0.8.conf
    ├── 10.10.0.8.log
    ├── 10.10.0.9.conf
    ├── 10.10.0.9.log
    └── start.sh

3 directories, 28 files

scenario.yaml controls all the configuration of benchmark. You can pass your own scenario by using -f option. By default, bgperf creates it automatically and places it under /tmp/bgperf like above. Let's see what's inside scenario.yaml.

$ cat /tmp/bgperf/scenario.yaml
<%
    import netaddr
    from itertools import islice

    it = netaddr.iter_iprange('100.0.0.0','160.0.0.0')

    def gen_paths(num):
        return list('{0}/32'.format(ip) for ip in islice(it, num))
%>
local_prefix: 10.10.0.0/24
monitor:
  as: 1001
  check-points: [1000]
  local-address: 10.10.0.2
  router-id: 10.10.0.2
target: {as: 1000, local-address: 10.10.0.1, router-id: 10.10.0.1}
testers:
- name: tester
  neighbors:
    10.10.0.10:
      as: 1010
      filter:
        in: &id001 []
      local-address: 10.10.0.10
      paths: ${gen_paths(100)}
      router-id: 10.10.0.10
    10.10.0.100:
      as: 1100
      filter:
        in: *id001
      local-address: 10.10.0.100
      paths: ${gen_paths(100)}
      router-id: 10.10.0.100
    10.10.0.101:
      as: 1101
      filter:
        in: *id001
      local-address: 10.10.0.101
      paths: ${gen_paths(100)}
      router-id: 10.10.0.101
...(snip)...

It describes local address, AS number and router-id of each cast. With regard to tester, it also describes the routes to advertise to the target.

check-points field of monitor control when to end the benchmark. During the benchmark, bgperf.py continuously checks how many routes monitor have got. Benchmark ends when the number of received routes gets equal to check-point value.

As you may notice, scenario.yaml is mako templated. You can use mako templating to simplify your scenario.