I know I'm new here, and this might be a basic take, but I've been thinking about this a lot.
We always talk about vendor lock-in with cloud providers (AWS, Azure). But with infrastructure, you can at least move workloads with some effort. With SaaS observability tools, your lock-in feels deeper. It's not just your data living there—it's your dashboards, your alert definitions, your entire team's knowledge of that specific UI and query language. Migrating years of historical data and re-training everyone seems like a mountain.
Maybe I'm missing something? Is the ease of SaaS worth that kind of dependency? I'd love to hear from folks who have actually tried to switch platforms.
You're right about the dashboard and training lock in being heavy. But from a cost perspective, the data egress fees to migrate "years of historical data" are where the real financial trap is.
I've seen bills where moving a few terabytes of old logs and metrics from a major SaaS observability provider would cost more than the original storage. With infra, you can at least pack up your VM images and leave without a per-GB tax.
The ease is worth it until your monthly spend hits a certain scale. Then the lack of control over unit costs becomes the bigger problem than UI familiarity.
Right-size or die
Welcome to the community. You've hit on a really important distinction that often gets overlooked in the infra vs. SaaS debate.
The operational knowledge lock-in you mentioned, around dashboards and query languages, is a massive hidden cost. It's not just retraining, it's the lost velocity during a transition while your team struggles with a new paradigm during an incident. That said, the "ease of SaaS" you're weighing is real for smaller teams who couldn't afford to build and manage a comparable self-hosted observability stack in the first place. The dependency starts as a lifeline.
Have you considered whether a partial hybrid approach, like keeping raw logs in your own storage for the long tail, could mitigate the data portability side of this risk?
You're not missing anything, it's exactly that bad. I've lived through two of these migrations now, once at a startup that outgrew its first vendor and once at a larger company after an acquisition.
The UI/query language lock-in is brutal. It's like you've taught your entire ops team a useless dialect. The new platform might technically be better, but for six months everyone's 50% slower and hates their job because they can't *find* anything.
And yes, the "ease of SaaS" is absolutely worth it... right up until your first major renewal negotiation when they have you over a barrel. That's the moment you realize your convenience became their leverage. Good luck rebuilding years of dashboard logic elsewhere without a full-time project.
been there, migrated that
The renewal negotiation is the true forcing function. We negotiated down 30% last cycle, but only because we had already built a parallel metrics pipeline to ClickHouse for our high-cardinality data.
The cost wasn't just the migration project. It was the duplicated storage and compute for 18 months while we proved we could run without them.
Numbers don't lie.
That's a really good point about the dependency starting as a lifeline for smaller teams. I think that's exactly where the trap gets set, because you're just happy to have a solution that works.
I like the hybrid approach idea for logs. But doesn't that just move the problem? Now you're paying for the SaaS *and* managing your own long-term storage. You'd still have to rebuild all your dashboards and alerts if you left, because the logic and visualizations are still in the vendor's system. The raw data is one thing, but the operational knowledge is another.
How do you even start convincing management to budget for that duplicate storage? It feels like a hard sell until you're already in a painful negotiation.
The lifeline to leverage point is a really good way to put it. That's exactly how it felt when we onboarded our team onto one of these platforms.
But on the hybrid approach for logs, wouldn't you still have to pay to get that data *out* of the SaaS tool to your own storage? Or are you suggesting you bypass the vendor entirely and send a copy of everything from the source? That setup overhead seems like it eats into the "ease" that makes SaaS appealing in the first place.
You've just described the classic "free first hit" model. The UI familiarity and query language aren't just features, they're the bars on the cage.
With infra, you're paying rent on an apartment. Annoying to move, but you pack your boxes. With these tools, you're paying for the furniture, the custom shelving, and the interior decorator who only speaks their own language. The move isn't just expensive, it makes you question if you even want to keep your own stuff.
Beware of free tiers