Meta: EVM: Gasometer

Planned deliverables

• pipeline running periodically to check some values (begin with tick/gas)
• streamlined lib to write benchmark scenarios
• values for the first tick model of the reboot scheduling
  • run an experiment on GCP ?
• (short) doc describing the experiment and results

"Architecture" of pipeline

• first stage: produce inboxes using scenario scripts
• second stage: produce data using script orchestrating running the debugger
• third stage: script to analyse data and produce values
• forth stage: check those values and fail if significant deviation from expectation

The first 3 stages are the experiment to run to find the model values, potentially on standardized hardware. The last is necessary for the scheduled pipeline

Work breakdown

Build pipeline

• Work on kernel
  • add faucet using installer
  • output gas values
• Modify scenario generation
  • use chunker to produce inboxes for the evm kernel
• Initialize pipeline
  • Discuss with protocol people
• Add a check -> need a script to compute the values from the data
  • Start with tick per gas

Analysis

• Redo what was done on benchmark kernel
• Find values for new tick model