Latest / The Cloud Business Podcast with Fexingo: AWS, Azure, GCP, and Enterprise Infrastructure / Why Cloud Contracts Now Meter Your Backup Snapshot Storage
Transcript
- Lucas: Luna, have you ever opened your cloud bill and found a line item that just didn't make sense? A charge that basically doubled overnight with no obvious change in what you're actually doing? Luna: More times than I'd like to admit. There's always some new metering dimension you didn't even know existed. Lucas: Right. And one that's been catching a lot of engineering teams off guard this year is how cloud providers are now charging for backup snapshots. Not the raw storage — that's been metered forever — but the metadata, the change tracking, the snapshot lifecycle itself. Luna: So it's not just 'how many gigabytes are you storing.' It's 'how many snapshots have you taken, how old are they, and how many incremental changes do they track.' Lucas: Exactly. And this is a shift that's quietly rolled out across AWS, Azure, and GCP over the past 12 to 18 months. Each provider took a slightly different approach, but the net effect is the same: if you're doing frequent incremental backups — say, every hour for a production database — your snapshot-related costs could be three to four times what they were two years ago. Luna: I've seen that firsthand. A friend at a mid-size SaaS company — about 200 employees, they run their app on AWS — told me their monthly backup bill went from around $1,200 to $4,800 in one quarter. They hadn't changed their backup policy. The only thing that changed was the pricing model. Lucas: That's exactly the kind of story that makes this worth unpacking. So let's start with AWS, because EBS snapshots are probably the most common enterprise backup mechanism in the cloud. For a long time, AWS charged you only for the actual data stored across all your snapshots. If you took a snapshot every hour and only changed 100 megabytes each time, you paid for that 100 megabytes, plus the baseline. Simple. Luna: But now they've introduced what they call a 'per-snapshot metadata fee.' Lucas: Right. Starting late last year, AWS began charging a small hourly fee per snapshot, regardless of its size. It's not huge — I think it's something like $0.0003 per snapshot per hour — but when you have hundreds or thousands of snapshots, that adds up fast. For that SaaS company, if they were keeping, say, 500 snapshots at any given time, that's an extra $360 a month just in metadata fees. Luna: And that's on top of the storage you're already paying for. The sneaky part is that the fee applies to every snapshot in your account, including ones you might have forgotten about — old dev volumes, pre-production backups, things that are just sitting there. Lucas: Precisely. Azure did something similar with their managed disk snapshots, but they attacked it from a different angle. Azure introduced a 'snapshot archive' tier that's cheaper per gigabyte but charges you a premium for accessing or deleting snapshots before a 90-day minimum retention period. The idea is to discourage short-lived backups and push you toward fewer, longer-lived snapshots. Luna: So it's a behavioral nudge, not just a fee. They want you to batch your snapshots into daily or weekly instead of hourly. Lucas: Exactly. And then GCP — they took yet another route. For their Compute Engine persistent disk snapshots, they've always charged for the actual data stored, but they recently tightened how they calculate 'unique data.' Previously, if you had two snapshots of the same disk, GCP only charged for the data that was different. But now they've clarified that if you delete an earlier snapshot, the data that was unique to that snapshot — but still referenced by later snapshots — can become 'orphaned' and you continue to be billed for it. It's a subtle change in the accounting that can lead to surprise charges when you clean up old snapshots. Luna: That's almost like a trap if you're trying to save money by deleting old backups. You delete what you think is redundant, and your bill actually stays the same or even goes up. Lucas: Yes. And that's counterintuitive. So the theme here is that cloud providers are moving from a simple 'pay for what you store' model to a more complex 'pay for how you store it' model. And this has real implications for how you architect your backup strategy and negotiate your next cloud contract. Luna: Speaking of which, I think this is a good moment to mention that if you're finding these kinds of insights useful for what you're building or running, listener support is what keeps this show independent and ad-free. You can help out at buy me a coffee dot com slash fexingo. It's a small way to keep the conversations going. Lucas: And we really appreciate that. Now, back to the backup pricing shift — let's talk about what you can actually do about it. The first move is auditing your snapshot inventory. Most teams have no idea how many snapshots they're retaining or what they're paying for them. Luna: And that's where tools like AWS Compute Optimizer or Azure Cost Management come in. You can set up reports that show snapshot costs by volume, by age, and by region. Lucas: Right. The second move is to rethink your snapshot frequency. Do you really need hourly backups for every volume? For critical production databases, maybe yes. But for dev or staging environments, daily or even weekly might be sufficient. And combining that with a lifecycle policy that automatically deletes snapshots older than, say, 30 days can dramatically reduce your metadata fee exposure. Luna: On AWS, you can use Amazon Data Lifecycle Manager to automate that. On Azure, you have Azure Policy and lifecycle management for blob storage. On GCP, you can set snapshot retention policies at the disk level. Lucas: Exactly. And the third move — this is the one I think most teams miss — is to consolidate your snapshots. Instead of taking individual snapshots of each volume, consider using a backup service that creates a single, consistent snapshot of your entire application stack. Services like AWS Backup or Azure Backup charge a flat fee per protected instance, and they handle the snapshot lifecycle for you, often more cheaply than doing it yourself. Luna: But that only helps if you're on a single cloud. For multi-cloud setups, you might need a third-party tool like Veeam or Commvault that abstracts away the provider-specific pricing. Lucas: Good point. And that brings us to the contract negotiation angle. When you're renewing your cloud commitment — say, a one-year or three-year AWS Enterprise Discount — you should ask specifically about snapshot pricing. Some providers are willing to include snapshot metadata fees in the discount calculation if you ask. But it's not automatic. Luna: So it's a negotiating point. And I've heard that if you can demonstrate you're a high-volume snapshot user, you might get a custom pricing tier. Lucas: Exactly. One more thing: watch out for snapshot replication costs. If you're copying snapshots across regions for disaster recovery, you're paying for the data transfer plus the storage in the destination region. And some providers now charge a separate fee for the replication service itself. AWS charges for cross-region snapshot copy, and Azure has a similar charge for geo-redundant storage. Luna: So the total cost of a single snapshot can include the source storage, metadata fees, replication transfer, destination storage, and maybe even an egress fee if you're copying to a different account. Lucas: Right. And that's why a simple change like reducing snapshot frequency or automating cleanup can have an outsized impact on your bill. I've seen teams cut their backup costs by 40% just by implementing a 90-day retention policy instead of indefinite retention. Luna: But you have to balance that against your recovery point objective. If you need to recover to within five minutes, you can't just snap every hour. Lucas: True. And that's where the trade-off lives. The cloud providers are essentially monetizing that granularity. The more granular your recovery needs, the more you pay in snapshot fees. So it's worth asking: do you actually need five-minute granularity, or would fifteen minutes be acceptable? Because the cost difference can be substantial. Luna: And I think that's the key takeaway. The cloud backup pricing model has shifted, and you need to actively manage it — not just set it and forget it. Lucas: Exactly. So the next time you open that cloud bill and see a line item you don't recognize, check whether it's a snapshot metadata fee. And then take a hard look at your backup policy. The savings might be hiding in plain sight.