lb-telemetry: Identifying optimisation opportunities in LHCb software tools

More time to gather data is required before optimisation analysis can begin, but many other useful results were obtained in the meantime. All statements below refer to internal CERN users only.

PIDCalib2: 

• Over the summer, the make eff hists module was invoked around 200 times per week • The sample is the biggest factor in the module's running time (the Turbo sample increasing it 10-fold on average)

• The module's execution time averages 7 mins and adds to 117 hours of total running time over the summer so it is well worth exploring optimisation strategies

LbConda: 

• There are around 8000 activations per week 

• The most popular configuration by far is: – OS: glibc 2.17 – Python: CPython 3.9.16 LbEnv:

• The /usr/bin/bash shell is by far the most popular, being used in over 99% of instances

 • Only 1% of users use a non-stable flavour (unstable or testing)

 • On average, there are about 10,000 total activations per day

• Broadwell and Skylake chips are about equally popular processors, but together are used in over 99% of all activations.