Hey everyone! We're considering a switch from using standalone Prometheus to Grafana Mimir, mainly to gain some performance and efficiency benefits. Right now, we have two separate Prometheus servers collecting metrics, and we're utilizing Promxy for a unified query experience.
If anyone has experience with this transition or can share useful blog posts or benchmarks comparing these two options, I'd love your insights—especially on:
* **Query Performance:** How does Mimir perform on long-range queries (like 6+ months) with its HA and MinIO backend in comparison to local Prometheus TSDB?
* **Storage Efficiency:** How does Mimir's storage usage stack up against local Prometheus storage for the same retention periods?
* **Quorum and Minimum Footprint:** Does Mimir need at least **3 hosts** for quorum or high availability? What's the smallest deployment you can feasibly use for HA?
Thanks in advance for your help!
4 Answers
Choosing between Mimir and Prometheus really comes down to architecture rather than just performance. Prometheus is a single-instance monolith, while Mimir is a microservices-based, scalable solution. Mimir might outperform Prometheus under certain conditions, but usually, the decision is based on the need for scalability rather than pure performance.
I’d suggest considering VictoriaMetrics. I’ve been using it, and it’s super performant and easy to set up. Plus, it offers the capability to push metrics, which is really handy for on-prem systems facing multiple firewalls. Gathering metrics gets tricky with too many ACLs!
Mimir is like what Prometheus would be if it turned into a Kubernetes startup. It can improve long-range queries and compression, but you trade simplicity for a bit of operational complexity that you might not be ready for just yet.
Mimir might not be as efficient due to its reliance on networking and object storage, which can slow things down. Prometheus typically uses in-memory caching, making it quicker for queries. When using Mimir, you have to scrape data, send it to a receiver, and create TSDB blocks before storing in object storage, which adds overhead. For long-term storage, using object storage can be cost-effective, but don't forget about those extra request costs. I often find myself recommending Thanos instead, as it lets you stick with Prometheus, which has its own advantages for scraping, storage, and querying.
Related Questions
Can't Load PhpMyadmin On After Server Update
Redirect www to non-www in Apache Conf
How To Check If Your SSL Cert Is SHA 1
Windows TrackPad Gestures