Skip to content

Latest commit

 

History

History
34 lines (22 loc) · 1.45 KB

BENCHMARKING.md

File metadata and controls

34 lines (22 loc) · 1.45 KB
Raw

Benchmarking LSKV

Distributed benchmarking

Requirements

The distributed benchmarking assumes a set of machines are available (running Ubuntu 20.04). Their IP addresses should be available in a hosts file one per line. The first node listed in this file will be the leader node and be responsible for launching the benchmarks from.

If using a VM scale set in Azure, running the following will get the IP addreses into the hosts file for you if they have public IP addresses:

az vmss list-instance-public-ips --name <vmss_name> --resource-group <vmss_rg> | jq -r '.[].ipAddress' | tee hosts

For VMs with private IP addresses reachable from a jumpbox you'll be running the commands from:

az vmss nic list --vmss-name <vmss_name> --resource-group <vmss_rg> | jq -r '.[].ipConfigurations[].privateIpAddress' | tee hosts

After provisioning the nodes, they need to be setup with the correct dependencies:

ansible-playbook -i hosts benchmark/distributed/setup_nodes.yaml -e @benchmark/distributed/values.yaml

Then, to run the benchmarks, ssh onto the first node (ssh <user>@$(head -n 1 hosts)), cd /tmp/lskv and run things from there (e.g. . .venv/bin/activate && python3 benchmark/distributed.py).

Receipt verification

The receipt verification benchmark (benchmark/receipt_verify.py) uses a hard-coded receipt so can be run standalone without a running datastore. It can be run with python benchmark/receipt_verify.py.