We deployed the new M-Lab platform to 1/3rd of the M-Lab fleet, and now we need to assess whether or not it is a performance regression, relative to the old platform. As long as we can be sure the performance of the new platform does not constitute a regression, then we can roll out the new platform and be confident that we have not made anything worse.
We found that no site saw worse performance on the new platform as compared to the old platform. As such, we are rolling ahead with the new M-Lab platform worldwide, with a small exception for our US sites, where we are keeping the old platform running for the duration of the FCC’s Measuring Broadband America Program quiet period in support of a long-term project partner.
In an effort to allow others to reproduce our work, we put everything in the following Jupyter notebook (also available for direct download: