Hey everyone! 👋 I've been deep in the weeds comparing pricing models for our observability stack (we're currently evaluating a switch), and I keep coming back to this thought: for our use case, **per-host pricing feels so much more predictable and easier to budget for than per-GB.**
I know per-GB is the dominant model for logs and traces these days, and I totally get the appeal of "pay for what you use." But when I run the numbers, our usage is *so* spiky. A deployment goes out, an error cascade happens, a marketing campaign drives unexpected traffic—boom, our GB volume for the month jumps 40%, and the bill gives me a heart attack. 😅 With per-host, I know exactly what we’re paying for each monitored server, container, or lambda function. It’s a fixed line item in the budget.
Here’s my breakdown of why per-host feels more stable *for us*:
* **Budgeting Simplicity:** I can multiply our number of production hosts by the unit cost. Done. No frantic dashboards watching ingest rates mid-month.
* **Spike Immunity:** We can have a noisy neighbor process or a bug causing verbose logging, and it doesn't immediately impact cost. It gives the engineering team breathing room to fix the issue without the finance team knocking on my door.
* **Alignment with Infrastructure Planning:** Adding cost is a conscious decision tied to scaling our infrastructure (adding a new service/host), not just to application behavior. This makes capacity planning conversations much clearer.
Of course, I see the other side! If you have very stable log volume or you're incredibly diligent about filtering and sampling, per-GB can be cheaper. And for massive, auto-scaling environments with hundreds of ephemeral containers, per-host can get tricky.
But I'm curious—am I missing something? Has anyone else found per-host pricing to be a sanity-saver for cost predictability? Or have you moved away from it and found per-GB to be *more* predictable in the long run? I'd love to hear your experiences and any gotchas I should consider.
Happy comparing
Totally agree on the budgeting simplicity. We made the same call when our log volume started swinging wildly between quarters.
But here's the twist we ran into: per-host made us lazy about log hygiene. When a cost spike doesn't hurt, it's easier to ignore that verbose debug logging left on in production, or that new microservice emitting a firehose of events. We ended up with a ton of noise in our data, which made finding real signals harder later. We had to self-impose some discipline and treat our ingest like we were still paying per-GB.
It's a great model for predictability, just don't let it kill your incentive to manage data quality
Stay curious.