Free tier load tests request
Hi again.
We're currently evaluating the possibility of a migration to Neon but we'd like to run some load tests but we currently use a free account. The type of tests that we're expecting aren't huge:
- 10 requests per second as normal load
- 120 as peak of traffic for a couple of minutes
- 300 for one minute for a stress test
Is this something we could run safely with our current plan?
3 Replies
ratty-blush•16mo ago
Hey đź‘‹
I think that this should be ok, there is an article that you might want to check out:
https://neon.tech/blog/autoscaling-in-action-postgres-load-testing-with-pgbenchOne thing to keep in mind with the free plan is that you have 5GB of egress transfer limit, so you might hit that with your load testing. It could be a good idea to plan an upgrade to not have to worry about that limit depending on the amount of tests that you are planning to run. You could also always reach out to the Neon sales team as well:
https://neon.tech/contact-sales
Neon
Autoscaling in Action: Postgres Load Testing with pgbench - Neon
In this article, I’ll show Neon autoscaling in action by running a load test using one of Postgres’ most popular benchmarking tool, pgbench. The test simulates 30 clients running a heavy query. While 30 doesn’t sound like a lot, the query involves a mathematical function with high computational overhead, which signals to the autoscaler-agent th...
Neon
Contact Sales — Neon
Interested in learning more about our plans and pricing? Contact our sales team.
foreign-sapphireOP•16mo ago
Thanks!
Thanks, so we actually upgraded the plan and did some testing.
CONDITIONS
We set up a Neon DB with 4 VCPUs and 16Gb RAM (the max that our plan allowed) without autoscaling
We ran a test with about 5 requests per second in 3.5 minutes, the max concurrency is 10 requests at a time
RESULTS
- 962 requests were made.
- We found that the average successful request took about 565ms
- 4,15% of requests failed, however my network wasn't good enough today and could also have introduced some failure in it.
Are these numbers that look fine to you? We honestly expected less latency before we started the tests
fascinating-indigo•16mo ago
Hey @igmoweb,
Can you share your script? I'd love to run the numbers for you.
Also, can you share the following:
- did you use pooled endpoint?
- which region was your testing service in?
- which region was your Neon database hosted in?
We set up a Neon DB with 4 VCPUs and 16Gb RAM (the max that our plan allowed) without autoscalingIs that the min CU size, or is that the upper limit for the range in your Neon instance?