Profiling Scripts

Scripts for running metric collections using the Recoder/Repalyer framework.

Usage:

run_profiling requires exactly 1 action and at least 1 metric

./run_profiling <action> <metrics>

For example, the following runs end-to-end profiling on two metrics: storage-update and evm-call. This includes running the replayer, extracting raw data, storing metrics to a database and generating metric graphs.

./run_profiling all storage-update evm-call

Actions

List of available actions

• record records substates and generate evm-call metrics data in a log file.
• replay replays substates and generate metrics data in a log file.
• extract extracts data from a log file and store in a database.
• plot plots graphs for specified metrics.
• all performs replay, extract and plot.

List of available metrics

• storage-update gets storage update size after each transaction
• code-size gets code size of smart contracts
• evm-call gets runtime of evm calls
• evm-opcode gets statistics of evm opcodes

Requirements

• sqlite3 database for storing metrics.
• gnuplot for plotting metric graphs.

Configurations

Before start the profiling script, parameters such as preferred file locations, block range and event file should be configured in ./run_profiling.

General configuration

DATABASE_FILE: Sqlite3 database location where metrics are stored. LOG_PATH: Output path for generated logs from substate-cli command. PNG_PATH: Output path for metric graphs.

The Replayer configuration

These parameters are required to execute replay command. WORKERS: Number of worker threads that execute substates in parallel. FIRST_BLOCK: The first block to replay transactions. LAST_BLOCK: The last block to replay transaction (inclusive range).

The Recorder configuration

These parameters are required to execute record command. EVENT_FILE: Fantom event files exported from Fantom chain via opera export. GENESIS: Fantom genesis file (optional). OPERA_DATADIR: Chaindata directory (optional).