benchmark-nodeos

Benchmark Runs of Nodeos under different configurations. Original purpose detail exSat Memory Requirements.

Overview

This repo, contains a several named configurations that are used to configure the Operating System, and Nodeos. With the named configuration Nodeos is started from a snapshot or genesis and run to a specific block height from block logs. During this replay lots of metrics from both the OS and Nodeos are collected to a file. At the end of the run some summary statistics will be reported. In addition the metrics file will contain the git commit hash and name of the configuration files used.

Initial

Host Hardware

It depends on provider. Needs to be provided in a file. /var/local/hwinfo.txt

System Configuration

Will collect

• start time date -u +"%Y-%m-%dT%H:%M:%SZ"
• total ram free -g | grep Mem: | xargs | cut -d" " -f2
• total swap swapon --show=Size --noheadings
• number of 1Gb pages grep HugePages_1048576kB /proc/meminfo
• OS Info uname -a
• Processor Model cat /proc/cpuinfo | grep 'name'| uniq
• Num Processor cat /proc/cpuinfo | grep process| wc -l

Nodeos Version

• full version of nodeos

Nodeos Config

• git URL to file and git commit tag

Running Metrics

Collected Every 60 seconds

System Metrics

• time stamp date -u +"%Y-%m-%dT%H:%M:%SZ"
• vmstat
  • procs running
  • memory swap, free, buff, cache
  • swap in and swap out
  • IO blocks in and blocks out
  • CPU user, system, idle, waiting for IO, stolen from VM

Nodeos Metrics

• head block
• LIB

DB Size Metrics

• free_bytes
• used_bytes
• reclaimable_bytes
• size

Parsing Output

The python script parse_stats.py get the latest files from /tmp/run and parses them. Both the Host Info and Stats files are reformed to JSON and Comma Separated Vales (CSV) The new files are written to the specified director and contain a unique id to link the files.