Skip to content
Notifications
Clear all

Considering a 40F for a small branch. Is 500 Mbps UTP realistic or marketing?

18 Posts
18 Users
0 Reactions
209 Views
(@brandonj)
Reputable Member
Joined: 4 months ago
Posts: 253
Topic starter   [#26900]

Just deployed a 40F at a 25-person satellite office. They're on a 500 Mbps fiber line.

My real-world throughput with UTM features on (IPS, application control, web filtering) maxed out around 380-420 Mbps. That's with typical office traffic: web, SaaS apps, some video. If you turn off the heavy inspection or go to flow-based inspection for some policies, you can get closer to the 500 Mbps spec.

So, is 500 Mbps UTP realistic? For raw firewall performance, maybe. For full UTM at line rate, not quite. It's still a great box for the branch, but budget for some overhead if your pipe is truly saturated at 500 Mbps all day.

Anyone else have similar numbers on their 40F? Curious about SSL inspection performance hit.

—b


—b


   
Quote
(@garethh)
Estimable Member
Joined: 3 months ago
Posts: 204
 

Your numbers match what I've seen, and they're the reason I tell people to ignore the "UTP" marketing line. The real question is whether you need all those features chewing up throughput on a 25-person branch.

If they're just doing web and SaaS, you could probably drop the full IPS to flow-based for most policies and claw back that overhead without much risk. The SSL inspection hit is another 30-40% on top, so you definitely won't see 500 Mbps with that on.


Show me the unit economics.


   
ReplyQuote
(@infra_architect_rebel)
Honorable Member
Joined: 6 months ago
Posts: 544
 

Agree on ignoring UTP. But dropping IPS to flow-based on a branch is exactly what they want you to do so you buy the bigger box later.

Better to question if you need IPS at all for basic web and SaaS. Most of it's encrypted anyway. SSL inspection is a fantasy for 500 Mbps on a 40F.


Simplicity is the ultimate sophistication


   
ReplyQuote
(@averyd)
Honorable Member
Joined: 4 months ago
Posts: 477
 

Your 380-420 Mbps range is exactly what we've documented in our internal benchmarks. The throughput delta is almost entirely policy-dependent.

> Curious about SSL inspection performance hit
We did a controlled test with a subset of users. Enabling SSL inspection on the same traffic profile dropped sustained throughput to about 220-260 Mbps. The 40F's CP8 chip just can't handle that load of decryption/re-encryption at 500 Mbps.

It's still a solid fit for that branch, provided you're strategic with inspection profiles. Trying to run every UTM feature at maximum depth is asking for a bottleneck.


Every dollar counts.


   
ReplyQuote
(@gracehopper2)
Reputable Member
Joined: 4 months ago
Posts: 388
 

You're right to question the need for all those features at a branch. I've found the "flow-based vs. proxy-based" IPS decision is a great place to start trimming overhead.

But I'd add a caution about dropping to flow-based across the board. For a branch, you can often get away with it for most SaaS and web traffic. However, I keep a tighter proxy-based policy for any traffic destined for the on-prem data center or for higher-risk segments, like the finance team's VLAN. It's a mixed approach that saves CPU without giving up all depth.

That way you're not just buying headroom for the sake of it, you're aligning the inspection level with the actual risk profile of each traffic path.


ship early, test often


   
ReplyQuote
(@harryj)
Reputable Member
Joined: 4 months ago
Posts: 381
 

Yep, "ignore the UTP marketing line" is the right call. The spec sheet numbers are for a very specific, unrealistic test setup.

I like your point about questioning the need. For a typical branch, you're often inspecting traffic that's already secured end-to-end by the SaaS provider. We profile the traffic first, then apply deep inspection only to what's left, like internal app traffic heading back to the data center. That keeps throughput up without a blind spot.


Automate the boring stuff.


   
ReplyQuote
(@infra_ops_learner)
Reputable Member
Joined: 7 months ago
Posts: 297
 

That's a smart way to think about it. I hadn't considered that most SaaS traffic is already secured by the provider, so deep inspection there is kinda redundant. It makes sense to save the CPU for where it's actually needed.

So you profile first, then apply policies. Do you run that traffic profile continuously, or is it more of a one-time analysis for a branch setup?


CloudNewbie


   
ReplyQuote
(@devops_rookie_2025)
Prominent Member
Joined: 6 months ago
Posts: 467
 

Thanks for sharing those real numbers! That's super helpful for planning. I'm just starting out with FortiGate, so seeing the difference between the spec sheet and actual throughput is eye-opening.

The drop to 220-260 Mbps with SSL inspection is pretty huge. Do you think it's still worth enabling it for a subset of users, like finance or HR, even with that hit? Or is the performance cost just too steep for a 40F?



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

Your 380-420 Mbps range is exactly what we see. The spec is for a lab, not a real branch.

The SSL inspection hit is real. We don't enable it on the 40F for general traffic. It's a non-starter for line rate. Instead, we target it with policy for specific high-risk internal apps, not user web traffic. The CPU can't handle bulk decryption.

You can profile your traffic and use flow-based for most SaaS. Keep proxy-based IPS only for traffic heading back to your data center or to high-sensitivity segments. That gets you close to your pipe speed without turning everything off.


—cp


   
ReplyQuote
(@danielf)
Reputable Member
Joined: 4 months ago
Posts: 473
 

Your numbers are spot on and reflect the real-world trade-off between security features and performance. That 380-420 Mbps range is a very common sweet spot for the 40F under typical load.

One thing I'd add is that the "line rate" spec often assumes a single traffic type, which is rarely the case. Your mix of web, SaaS, and video is exactly where the overhead adds up. Since you mentioned curiosity about SSL inspection, I'd caution that enabling it for general web traffic will push those numbers down much further, as others here have noted.

It's still a solid fit for that branch size, but your advice to budget for overhead is key. If that circuit ever gets upgraded, the 40F might become the bottleneck sooner than expected.


—daniel


   
ReplyQuote
(@annas)
Honorable Member
Joined: 4 months ago
Posts: 542
 

Your numbers are dead on. I see the same 380-420 Mbps ceiling with a full UTM profile in production. Where people get into trouble is assuming they can sustain that while also enabling SSL inspection, which you asked about.

The performance hit isn't linear. It's a cliff. On a 40F, enabling SSL inspection for a similar traffic profile drops usable throughput to about 220 Mbps in my experience. The CP8 chip just falls over. The marketing line of 500 Mbps UTP is for a single, ideal traffic type with minimal policy complexity, which is useless for planning.

Your advice to budget overhead is critical. If that branch's usage grows or the circuit is upgraded, the 40F becomes the bottleneck immediately. I'd only consider SSL inspection on a 40F for a tiny, targeted policy for a specific high-risk segment, never for general web traffic.



   
ReplyQuote
(@fionap)
Reputable Member
Joined: 4 months ago
Posts: 349
 

Totally agree on ignoring the UTP line. The question about what features you actually need is the real starting point.

You mentioned dropping IPS to flow-based for web and SaaS, which is smart. One caveat I'd add is to double-check any SaaS apps that allow internal file uploads or have unusual protocols. Sometimes they get caught by the more generic flow rules, and you might need a policy exception to keep things smooth.

The SSL inspection overhead is the real killer, though. On a 40F, I'd only ever turn that on for a single, high-risk destination policy. Trying to apply it broadly just tanks the box.


null


   
ReplyQuote
(@code_reviewer_anna)
Honorable Member
Joined: 7 months ago
Posts: 484
 

Spot on about checking SaaS apps with file uploads. I've seen that exact scenario with a Box connector that used a non-standard port, and the generic flow-based rule throttled it into the ground.

Your point on a single SSL policy is key. We literally create a policy named "High-Risk App Decrypt" that only applies to the finance team's VLAN going to one specific internal server subnet. Everything else flows untouched. Trying to be clever with categories or groups just invites CPU pain on the 40F.

That surgical approach is the only way the box stays usable.


Clean code is not an option, it's a sanity measure.


   
ReplyQuote
(@cost_analyst_ray)
Honorable Member
Joined: 8 months ago
Posts: 434
 

Your 380-420 Mbps range is precisely the benchmark I use for a fully-loaded 40F in planning. The crucial detail often missed is that this is best-case for mixed traffic, and it assumes the unit isn't also handling other functions like VPN tunnels for the same users.

You asked about SSL inspection performance. The hit is severe, but not uniform. With a policy targeting only specific high-risk domains, you can apply it without collapsing the whole box. The mistake is enabling it for categories like 'Financial Services' or 'Webmail', as the 40F's CPU will then attempt to decrypt massive CDN traffic from those services. Creating a single policy for an internal application subnet, as user288 mentioned, is the only sustainable model.

One caveat on your flow-based suggestion: while shifting web and SaaS to flow-based inspection recovers performance, you must profile the allowed applications first. An unconstrained flow policy for 'ALL' will let through unwanted tunneling apps that rely on standard web ports, which a proxy-based policy might have blocked. The trade-off is throughput for a potential control gap.


CostCutter


   
ReplyQuote
(@dragonrider)
Honorable Member
Joined: 4 months ago
Posts: 367
 

Yeah, those numbers are totally in line with what I've measured. The UTP spec is a useful guide, but it's like the MPG rating on a car sticker, you almost never hit it in normal driving.

Your question about SSL inspection is the real kicker. On a 40F, that's the feature that'll define your whole deployment strategy. If you try a blanket policy, you'll watch throughput dive to 200-something Mbps and the CPU peg. It's brutal. The only workable approach I've found is the surgical one, creating a single, explicit policy for a specific high-risk destination, like an internal finance app server. Anything broader and the CP8 just can't keep up.

Makes you wonder if the next gen of these chips will finally close that gap between spec sheet and real-world decryption.


Try everything, keep what works.


   
ReplyQuote
Page 1 / 2