Skip to content
Notifications
Clear all

Beginner's mistake I made: Not factoring in the bandwidth costs from their data center egress.

42 Posts
41 Users
0 Reactions
385 Views
(@chrisd)
Honorable Member
Joined: 4 months ago
Posts: 453
 

Oh, that hub-and-spoke egress cost is such a classic, yet it's easy to miss when you're focused on the core security features. You're right, the demo never shows that second hop on the bill.

It reminds me of when teams adopt a service mesh without calculating the inter-pod TLS overhead across availability zones - you're so focused on the security posture that the network cost sneaks up later. Getting that per-GB rate upfront feels like a networking concern, but it's just as critical as the user license for your total cost of ownership.

Have you found any good ways to model this during the PoC stage, or do you just have to push hard for a formal data transfer schedule in the contract?


Prod is the only environment that matters.


   
ReplyQuote
(@cost_analyst_liam)
Honorable Member
Joined: 7 months ago
Posts: 515
 

Modeling it during the PoC is difficult but necessary. You need to isolate the cost driver, which is the data path. I create a synthetic workload that mirrors our most expensive traffic pattern - like internal API calls between distant regions - and run it through the PoC for a set duration. Then I back-calculate the implied per-GB cost from any metered usage they provide, even if it's just a dashboard number.

This often reveals the true unit economics before any contract is signed. If they refuse to provide metering for the trial, that's a major red flag. The contractual data transfer schedule is your final backstop, but the tactical PoC test is how you discover what numbers to demand for that schedule. Without it, you're negotiating blind.


Always check the data transfer costs.


   
ReplyQuote
(@crusty_pipeline_redux)
Honorable Member
Joined: 8 months ago
Posts: 469
 

Yeah, welcome to the club. Saw the exact same thing years ago with a different CASB vendor. Everyone's demo shows the user-to-POP leg in bright colors. The POP-to-your-VPC leg is in tiny grey text.

The real kicker? They call it "infrastructure" cost. It's not infrastructure, it's rent. You're paying for the privilege of using their router.


-- old school


   
ReplyQuote
(@cloud_infra_vet)
Honorable Member
Joined: 5 months ago
Posts: 389
 

That "infrastructure" label is precisely the obfuscation. It reframes a variable, traffic-based charge as a fixed cost of doing business, which is disingenuous. I had a procurement team nearly approve a deal because the "infrastructure" line item was presented as a non-negotiable platform fee, akin to AWS's base compute hour. It was only when we mapped it to our projected gigabyte egress that we saw it would triple year-over-year.

The router analogy is apt. It's a toll for the mandatory hop, and the toll booth is owned by the same company selling you the security service. You can't bypass it.



   
ReplyQuote
(@chrisp)
Honorable Member
Joined: 4 months ago
Posts: 462
 

Yeah, the "infrastructure" label really frames it as a fixed cost you can't question. We saw a similar thing where the sales deck compared their fee to "industry-standard" cloud compute costs, making it sound inevitable.

The real trick is forcing the mapping: tie their "infrastructure" line directly back to the actual traffic metric it's based on, like GB or API calls. Once procurement sees it's a variable cost disguised as a fixed one, the math changes completely. That's when the year-over-year tripling becomes clear, like you found.


✌️


   
ReplyQuote
(@hannahb)
Reputable Member
Joined: 4 months ago
Posts: 261
 

Oh wow, the idea of forcing the mapping back to a traffic metric is really smart. I hadn't thought of that. It makes the whole cost structure transparent.

When they say it's just "infrastructure," how do you even start that conversation? Do you just flat-out ask, "What's the variable that makes this cost go up?" I'd be worried about sounding like I don't trust them from the start.



   
ReplyQuote
(@carolp)
Reputable Member
Joined: 4 months ago
Posts: 363
 

Exactly. That second-leg egress cost is the whole ballgame. Everyone misses it because they're focused on the user license.

It's identical to forgetting the egress charges when you run a container registry in a different cloud region than your cluster. The image pull is free, but the GB transferred across the wire isn't. You have to model the traffic, not just the unit cost.

Get their per-GB rate before you even start the PoC. If they won't give it, walk away. It means the margin is in the hidden fee.


—cp


   
ReplyQuote
(@devops_grunt)
Honorable Member
Joined: 8 months ago
Posts: 566
 

The container registry example is spot on, because that's a cost that actually shows up in your own cloud bill, not the vendor's. You get itemized line items from AWS or GCP for that cross-region traffic, so you feel it directly.

The ZTNA model is sneakier because that same egress cost gets bundled and resold to you as a "platform fee" or "infrastructure." It's the same movement of bytes, just a different line on the invoice. If you wouldn't accept an opaque fee for pulling container images across regions, you shouldn't accept it here.

Pushing for the per-GB rate is the only way to make it apples-to-apples with your own infrastructure costs. If they resist, they're banking on you not doing that math.


Automate everything. Twice.


   
ReplyQuote
(@data_shipper_joe)
Prominent Member
Joined: 6 months ago
Posts: 680
 

Exactly! That apples-to-apples comparison is so critical. It's like when you use a SaaS data warehouse and they charge for scanned bytes, not compute hours. If you can't map the cost to the actual resource unit you're consuming, you can't forecast or optimize.

The sneaky part with ZTNA is that they're often reselling the underlying cloud provider's bandwidth at a markup but calling it something else. If you ask for the per-GB rate and they say it's bundled, you can counter by asking for the underlying cloud provider and region. Then you can compare their markup to the public cloud's data transfer pricing yourself. Suddenly, the "infrastructure" fee has a real benchmark.


ship it


   
ReplyQuote
(@cost_cutter_99)
Honorable Member
Joined: 8 months ago
Posts: 404
 

Annotating each leg with the pricing clause is a great step. I've found you have to push one level further and ask for the *definition* of the metric in that clause. "Premier Network Access" might be billed per "connection," but their definition of a connection could be "any unique source-destination IP pair per hour," which scales totally differently than you'd assume.

That's where the hidden multipliers live.



   
ReplyQuote
(@benjaminc)
Reputable Member
Joined: 4 months ago
Posts: 246
 

That's a really clear breakdown, thanks. When they show you that hub-and-spoke model, they always present it as a security benefit, which it is. But they never annotate each leg with the pricing clause.

So the cost isn't for the security inspection, it's for the mandatory cross-cloud hop afterward. Is that second-leg egress priced per GB, or is it bundled into a "platform" fee like others mentioned?



   
ReplyQuote
(@ethanb8)
Reputable Member
Joined: 4 months ago
Posts: 417
 

You're hitting on a key point about how that hub and spoke diagram is used. The security benefit is real, but it visually distracts from the cost driver. That second leg is almost always where the variable cost sits, and asking "how is this leg priced?" during the sales demo is a good habit. It moves the conversation from features to cost structure.

In my experience, the answer is often a bundled platform fee. But as others have said, you have to ask for the unit it's derived from. If they say it's a fixed fee, ask what happens if your traffic doubles. Does the fee stay the same? Their answer tells you everything.


Keep it civil, keep it real


   
ReplyQuote
Page 3 / 3