Latest / The Cloud Business Podcast with Fexingo: AWS, Azure, GCP, and Enterprise Infrastructure / Why Cloud Bills Now Charge for Instance Scheduling
Transcript
- Lucas: So I was looking at a client's AWS bill last week, and there it was — a line item I'd never seen before. It said 'instance scheduling'. And it wasn't a one-time setup fee. It was metered, like every other resource. Luna: Wait, they're charging for the scheduler itself now? I thought that was just a free feature you toggled on. Lucas: That was the old model. And honestly, it's a natural extension of what we've been seeing all year — providers moving from bundled to metered pricing. We talked about NAT gateway processing, firewall rules, DNS queries. Now it's the scheduler. The thing that decides when your instances run. Luna: So how does it actually work? Is it per instance, per schedule, per action? Lucas: From what I've seen, it's metered per schedule execution. So every time your scheduler triggers a start or stop action on an instance, there's a small charge. And those add up fast if you're doing granular schedules — like starting and stopping hundreds of instances every few minutes for batch processing. Luna: I can see that. But for a typical dev environment, you might only schedule a start at 9 AM and a stop at 6 PM. That's two actions per day per instance. That can't be that expensive. Lucas: Right, per instance it's tiny. But when you're an enterprise with thousands of instances, and you're scheduling them across time zones and regions, the volume of actions gets big. And the pricing is per action, so it scales linearly. Plus, the providers are starting to offer 'premium scheduling' features — like predictive start based on historical usage — at a higher rate. Luna: So they're basically monetizing the convenience of automation. The thing that was supposed to save you money on idle compute is now a line item itself. Lucas: Exactly. And the interesting part is how it changes the cost-benefit calculation. Before, you'd set up a schedule to shut down non-production instances at night and save maybe 40 percent on those resources. Now you have to factor in the scheduling fees. For most companies, it's still a net win — the savings dwarf the fees. But for edge cases, like very small instances or very short-lived environments, the math can flip. Luna: Do you have a specific example? Because I feel like this is one of those things that's easy to hand-wave. Lucas: Sure. I talked to a retail company that runs batch jobs every 15 minutes during business hours. They have about 500 instances that they start and stop per batch. That's a thousand scheduling actions per cycle, four times an hour. So that's four thousand actions per hour, say 12 hours a day — that's 48,000 actions a day. At a tenth of a cent per action, that's 48 dollars a day. Over a month, that's about 1,500 dollars just for scheduling. Not huge, but not nothing. Luna: That's a real number. And if you're running 24/7, it doubles. So suddenly the scheduler is a meaningful cost center. Lucas: And that's just AWS. Azure has a similar model now, and GCP is rumored to be rolling out their own metered scheduling later this year. It's a trend, not an outlier. Luna: So what do you tell clients who are hitting this? Is there a way to optimize away the fees? Lucas: The first thing is to audit your schedules. Most companies have way more schedules than they need. We found one client running a schedule for a single instance that was already terminated. They'd been paying for a ghost schedule for months. So cleanup is step one. Luna: Ghost schedules — that's a good term. I bet there are a lot of those out there. Lucas: The second thing is consolidation. Instead of having separate schedules for each instance, you can group them into a single schedule that starts and stops a fleet. That reduces the number of actions, because the scheduler treats the group as one action. Some providers charge per group action rather than per instance. That can cut your costs significantly. Luna: So the granularity of your scheduling directly impacts the bill. The more fine-grained, the more you pay. Lucas: Right. And that's kind of the theme of this whole billing evolution — providers are monetizing granularity. Every time they give you more control, they meter it. And they're charging for the control itself. Luna: But is there a point where you just say, forget the scheduler, I'll handle it manually? Or use a third-party tool? Lucas: That's the workaround some teams are exploring. There are open-source schedulers and third-party tools that run on a single small instance and manage your fleet via APIs. You pay for that one instance and avoid the per-action fees. But then you have to maintain it, and you lose the native integration. For a lot of companies, the native scheduler is still worth it — it's more reliable and less work. Luna: So it's a trade-off between convenience and cost. That's a classic infrastructure decision. Lucas: Exactly. And there's a third option: move to serverless. If you're scheduling short-lived tasks, you can often run them on functions instead of instances. Functions are billed per execution, and they don't need a scheduler at all — the cloud provider triggers them. So you eliminate the scheduling fee entirely. Luna: But serverless has its own pricing quirks. We've talked about that before — the token metering, the GPU memory charges. It's not a free lunch. Lucas: Not at all. But for the right workloads, it can be cheaper overall. And it's a good reminder that the scheduler fee is just one line item in a much larger cost landscape. The real skill is seeing the whole picture. Luna: So if someone's staring at their bill right now, what's the one thing they should do today? Lucas: Audit your schedules. Look for any schedule that's not actively saving you money. If a schedule is running but the instances it manages are always on anyway, or if it's scheduling instances that are already stopped, kill it. That's free savings. And then look at your scheduling frequency — if you're scheduling every minute but your workloads don't need that, you're paying for nothing. Luna: That sounds like a great first step. And it's something you can do in an afternoon. Lucas: Yeah, it's a quick win. And honestly, the fact that we're even talking about this — that scheduling fees are a thing — tells you how deep the metering rabbit hole goes. Every time we think we've seen it all, there's a new line item. Luna: It's like the cloud providers are playing a game of 'find the hidden cost' with us. Lucas: And you know, that's part of why we do this show. We're here to help you spot those hidden costs before they become a surprise on your invoice. If you're an operator or a builder, and you're getting value out of these conversations, that's what keeps us going. If today was worth a coffee to you, you can buy us one at buy me a coffee dot com slash fexingo. Luna: Yeah, every little bit helps us keep the show ad-free and independent. And we genuinely appreciate the support. Lucas: So back to schedules — the other thing to watch for is the 'premium' features. Some providers are starting to offer machine learning–based scheduling that predicts when you'll need capacity. That's billed at a higher rate, but it can save you more in the long run if it reduces idle time. It's a bet on the optimizer. Luna: A bet on the optimizer — I like that. But how do you know if it's worth it? Is there a way to test it? Lucas: You can run a pilot. Take a subset of your instances, enable predictive scheduling, and compare the costs to your baseline. Over a month, you'll see if the savings in compute outweigh the extra scheduling fees. It's a data-driven decision, which is the right way to approach these new line items. Luna: So the takeaway is: don't panic, but do audit. And don't assume the scheduler is free anymore. Lucas: Exactly. And as the providers keep rolling out these metered features, the companies that stay ahead are the ones that treat their cloud bill like a living document — something they review and optimize regularly, not just a number they pay at the end of the month. Luna: That's a good note to end on. So what's next? Are you going to start charging us for scheduling episodes? Lucas: Ha, not yet. But if they start metering podcast production, we'll be the first to cover it.