Time2: InfluxDB, TimescaleDB Benchmarking

Group 6

Team Members
Natalia Bourdi	el19031
Nikolas Stamatopoulos	el19020
George Mystriotis	el19065

---

Overview

This project aims to compare the performance of two leading time-series databases, TimescaleDB and InfluxDB, using the Time Series Benchmarking Suite (TSBS). We benchmark both single and multi-node configurations, considering three different data sizes and two distinct use cases/database schemas.

The experiments were conducted using Docker containers on Ubuntu Desktop Server 16.04 running on 8 CPU cores @ 2.30GHz / 16GB RAM / 90GB Storage.

---

What was measured

• Load / insertion rates for datasets differing in size, partitioning, chunk times, and worker fleet (process parallelism)
• Query execution rates for various query types (point queries, aggregation queries, etc.), on datasets varying in partitioning

for the below configurations

• Single-node TimescaleDB
• Multi-node TimescaleDB (consisting of 1 access node and 3 data-bearing nodes)
• Single-node InfluxDB v1 (uses FluxQL)
• Single-node InfluxDB v2 (uses Flux)

^{InfluxDB multi-node was not tested as it is exclusive to the Enterprise plan.}

---

Setup and Requirements

Prerequisites: go, Docker, docker-compose.

1. Clone the repository

   git clone https://github.com/nbourdi/timescale-v-influx

   ^{All TSBS go programs useful for the application are included. For compatibility with Influx v2, some of the programs were tweaked.}

2. Build all tsbs executables/commands

   Navigate to the /tsbs directory and execute:

   $ make 

3. Setup and benchmark your desired database configuration(s)

   Use the relevant how-tos in docs.

---

Benchmarking Configurations

Images/Versions

• TimescaleDB 2.13 | PostgreSQL 14

This is the last version that supports multi-node.

• InfluxDB v2.7

The latest version of InfluxDB, using the newer Flux query language.

• InfluxDB v1.8

An older version which uses the SQL-like InfluxQL and is supported by TSBS.

---

Datasets generated/used

Both use cases implemented by TSBS were used for the benchmarking process.

1. IoT Data from Trucks:

   Size	Duration	Scale
   500MB	2 days	500 trucks
   5GB	20 days	500 trucks
   25GB	3 months	500 trucks

2. DevOps Readings:

   Size	Duration	Scale
   500MB	10 days	50 servers
   5GB	20 days	50 servers
   25GB	3 months	50 servers

---