Cost benchmarking - Self hosting SigNoz vs using a SaaS vendor
In this post we want to lay out typical price which someone would incur in running SigNoz. This would give potential users an idea of what resources they would need to provision & typical monthly cost at different application load and sampling rates.
In this post we would benchmark the cost of self hosting SigNoz and the price you will pay to SaaS APM vendors. For benchmarking, we will take DataDog as an example - as it is one of the more popular APM tools in the market.
This would give potential users an idea of what resources they would need to provision & typical monthly cost at different application load and sampling rates.
DataDog’s APM pricing (as on 8 Feb 2021)
USD 31 per host per month
- 15-minute Live Search & Analytics (150GB incl.)
- 15-day Historical Search & Analytics (1M Indexed Spans incl.)
- 15-month metric retention
- 1.7 USD/mn spans for 15 day retention
To compare the cost of running DataDog APM & SigNoz, we will take 15 day retention as the standard - as that is what DataDog uses as default.
We will take 90 RPS workload as an example to compare cost of running DataDog and SigNoz. We are creating more detailed benchmark frameworks. This is just a first attempt at this.
Please, note that 90 RPS is the ingested and retained trace rate. Many companies sample trace extensively ( upto 0.5%) but here we are comparing trace volume sent RPS. For example, if you are sampling traces at 1%, then you can handle upto 9000 RPS with below costs - both for SigNoz & DataDog.
DataDog APM cost
Cost of running SigNoz
For running this on SigNoz, we tested it on a EKS cluster of 3 nodes with 8GB RAM (t3.large). Monthly cost for t3.large (8 GB RAM, 2 CPU) is ~42 USD/ month ( with 30% annual reserved instance discount)
Total SigNoz cost ~ 338 USD per month
Although this setup works at less than 60% utilisation, some users may still want to have redundancy.
At least 10x cost improvement over DataDog
For 2x redundancy, cost for running SigNoz = 676 USD per month
Improvement over DataDog = 19982/676 = 29x improvement over DataDog's cost. Even if we add more level of over provisioning which different devops teams may want to add, we expect at least 10x improvement of cost over DataDog at this scale.
And finally, this is how it looks on your SigNoz dashboard 🤓
