This is the experimental evaluation code repo for the paper "Enabling Scalable Coordination Services for Metadata Management using Distributed Key-Value Databases: A Design Space Exploration".
Some scripts for setting up a benchmark environment for FDBKeeper.
pipenv install
-
Prepare at least 1 host for client, at least 1 host for foundationdb or zookeeper service.
-
Server requirements
- Ubuntu 22.04 is recommended
- Mount ext4 fs on /var/lib/keeper-bench with options:
defaults,noatime,discard
- Setup prometheus node exporter with textfile collector for
/var/lib/node_exporter
- Setup prometheus pushgateway for
keeper-bench
.
-
Create ansible inventory file:
hosts.yml
. Put hosts inclients
andservers
groups. Here is a simple example:all: children: clients: hosts: c1: ansible_host: 1.2.3.4 c2: servers: hosts: s1: fdb_coordinator: true fdb_procs: [transaction, storage] s2: vars: keeper_bench_path: /local/path/to/keeper-bench fdb_init_command: configure new triple ssd
We use prometheus and grafana to collecting metrics. You can use docker compose to deploy a simple prometheus:
# Create file sd configs if you setup hosts without `setup-ecnu-cloud` playbook.
touch docker/prom_file_sd.yml
# Up docker compose services
docker compose up -d
# Grafana is available on 0.0.0.0:9033
Open 127.0.0.1:9033 and setup data source and dashboard manually. Here is a dashboad example
See Makefile.
If there are any questions, please feel free to propose new features by opening an issue or contact with the author: Jun-Peng Zhu(zjp.dase@stu.ecnu.edu.cn). Enjoy the code.