LLAMA
L.L.A.M.A. is a deployable service which artificially produces traffic for measuring network performance between endpoints.
LLAMA uses UDP socket level operations to support multiple QoS classes. UDP datagrams are fast, efficient, and will hash across ECMP paths in large networks to uncover faults and erring interfaces. LLAMA is written in pure Python for maintainability.
Okay, but not yet - Alpha Status
LLAMA will eventually have all those capabilities, but not yet. For
instance, there it does not currently provide QOS functionality,
but will send test traffic using hping3
or a built-in UDP library
It’s currently being tested in Alpha at Dropbox through experimental
correlation.
Documentation
- LLAMA on ReadTheDocs
- Changelog
- See Issues for TODOs and Bugs
Visualization
Using InfluxDB and interested in visualizing LLAMA data in a matrix-like UI?
Check out https://github.com/dropbox/grallama-panel
Acknowledgements / References
- Inspired by: https://www.youtube.com/watch?v=N0lZrJVdI9A * with slides: https://www.nanog.org/sites/default/files/Lapukhov_Move_Fast_Unbreak.pdf
- Concepts borrowed from: https://github.com/facebook/UdpPinger/