A tool to produce statistics from Wireshark or tshark dumps.
cargo run --bin egui --release
# OR
cd analyser-gui
cargo run --releaseThis program will load and monitor a folder called ./dumps relative to its crate root (or
binary if running standalone). Put Wireshark .pcapng files in that folder and they'll show up in
the GUI for graphing.
The Postgres method should still work, but it's not efficient. Use the new custom GUI tool instead.
cargo run --release -- --cycle-ops 6 ./dumps/baseline.pcapngNOTE: --cycle-ops is ignored. TODO: Remove arg lol
sudo apt install -y postgresql-client
docker-compose up -d postgres adminerPostgres on port 5432 and adminer on 8080.
Import some data with:
./ingest ./dumps/baseline.pcapngImport all dumps with
./process-all.shThis seems to be the best, tied with Zeppelin. SQL can be written to show latencies and histograms and stuff.
dc up -d redash- Run this:
dc run --rm redash create_db - https://localhost:5000
- Make a user with generally any old garbage as the user/email/password, as long as you remember it.
- Make a Postgres data source with host/user/pass set to
ethercrab.
select
scenario,
(delta_time_ns :: float / 1000) as delta_time,
ROW_NUMBER() OVER (
partition by scenario
order by
packet_number asc
) counter
from
ethercrab
where
command like '%LRW%'
and scenario like '%merge_integrated%' --or scenario like '%new4card%'
or scenario like '%ecat-enc-48-49-i350-0usecs-i7-3770-netbenc%'
or scenario like 'ecat-enc-48-49-i350-0usecs-tadm-net-ltcy-i7-3770-netbench'
order by
counter asc
limit
10000SELECT
width_bucket(delta_time_ns, 0, 100000, 200) as bucket,
count(delta_time_ns),
int8range(min(delta_time_ns), max(delta_time_ns), '[]') as range,
scenario,
min(delta_time_ns) as min,
max(delta_time_ns) as max
from ethercrab
where scenario like '%merge_integrated%'
GROUP BY bucket, scenario
union
SELECT
width_bucket(delta_time_ns, 0, 100000, 200) as bucket,
count(delta_time_ns),
int8range(min(delta_time_ns), max(delta_time_ns), '[]') as range,
scenario,
min(delta_time_ns) as min,
max(delta_time_ns) as max
from ethercrab
where scenario like '%ecat-enc-48-49-i350-0usecs-i7-3770-netbenc%'
GROUP BY bucket, scenario
union
SELECT
width_bucket(delta_time_ns, 0, 100000, 200) as bucket,
count(delta_time_ns),
int8range(min(delta_time_ns), max(delta_time_ns), '[]') as range,
scenario,
min(delta_time_ns) as min,
max(delta_time_ns) as max
from ethercrab
where scenario like 'ecat-enc-48-49-i350-0usecs-tadm-net-ltcy-i7-3770-netbench'
GROUP BY bucket, scenarioIf anyone knows a better way of doing multiple overlapping histograms than just a UNION I'm all
ears.
Grafana starts up on port 3000 but is rubbish for non-time-series stuff, which is a shame. It's left
in the docker-compose.yaml for posterity.
- Start it with
dc up zeppelin. It will fail to start because of permissions errors. sudo chown -R 1000:1000 data/zeppelindc up -dshould work now
Zeppelin is a pretty good option for analysis
There's a Jupyter Lab image in docker-compose with the Rust stuff preinstalled. I haven't used
this yet, but it might be useful to get more customisable charts using plotters.