Firewalls are often chosen by looking at the biggest throughput number on the datasheet. The device is installed, IPS and SSL inspection are enabled, and even though the internet line is the same, users start complaining about slowness. The problem isn’t a faulty device but sizing based on the wrong number. The real performance of an enterprise firewall depends on the security features to be enabled, the traffic profile and the management model. The questions below provide a framework for a vendor-neutral comparison.
1. Define the need first
- Bandwidth and growth: Current internet lines and expected growth over 3–5 years.
- Security features to be enabled: IPS, application control, web filtering, antivirus and SSL inspection.
- Remote access: Concurrent VPN users, number of site-to-site tunnels, and MFA and ZTNA needs.
- Segmentation: Whether internal (east-west) traffic will also pass through the firewall; this decision directly affects capacity.
- Availability: Requirements for high availability (HA), dual lines and hardware redundancy.
Segmentation is the item most often underestimated. When the firewall only protects internet egress, it’s enough for it to carry a 1 Gbps line. If traffic between user VLANs and the server VLAN will also pass through the firewall, file sharing, backup and database traffic come into the calculation. Worked example: when internal traffic reaching 2–3 Gbps during business hours is added to 1 Gbps of internet traffic, the need multiplies several times. If this traffic will be inspected with security profiles, the number to measure isn’t firewall throughput but the threat protection value. Show the physical network counterpart of the access zones on the firewall in the enterprise VLAN setup and network segmentation plan.
Add the physical interfaces to the list too: how many 1G and 10G ports, SFP/SFP+ slots, ports reserved for HA heartbeat and lines to be added in the future. Putting a switch in between later because there weren’t enough ports breaks the design.
2. Read the datasheet numbers correctly
Vendors publish different numbers for different test conditions. There can be a manyfold difference between the “firewall throughput” measured with security features off and the “threat protection” or similar value measured with IPS, application control and antivirus on. With SSL inspection on, capacity drops even further.
| Metric | What does it measure? | Use in selection |
|---|---|---|
| Firewall throughput | Stateful tracking only, large packets | Not enough on its own to make a decision |
| IPS / threat protection | Realistic traffic with security profiles on | The main sizing criterion |
| SSL inspection | Inspection of encrypted traffic | Decisive if SSL inspection will be used |
| Concurrent / new sessions | Connection table and session setup rate | Important with many users and heavy web traffic |
| IPsec VPN throughput | Encrypted tunnel capacity | Important with heavy branch and remote access use |
The packet size and traffic mix used to measure a number are written in the datasheet’s footnote. A value measured with large UDP packets isn’t repeated in real office traffic made up of short HTTPS sessions. Before putting different vendors’ numbers side by side, match the test conditions; two numbers under the same heading may not measure the same thing. The RFC 9411 methodology published by the IETF for performance testing of network security devices can be used as a reference when questioning test conditions.
For models you’re undecided between, the most reliable method is to test with your own traffic. Most vendors and distributors provide trial units. Put the device in front of real traffic for a few days and monitor CPU usage and latency with all the profiles you’ll enable turned on. The datasheet is an estimate. Your own traffic is a measurement.
3. Licensing and total cost of ownership
Most security features run on annual subscriptions; when the subscription ends, signature updates stop. Compare not on hardware price but on the total of hardware + subscription + support over 3 or 5 years. The license requirement for the second device in an HA cluster varies by vendor.
When comparing offers, match the contents of each bundle line by line. Web filtering or sandboxing included in one vendor’s base bundle may be a separate license with another. Central management and logging products are also often priced separately. Also take into account that the renewal price doesn’t include the initial purchase discount. The real cost shows up at renewal.
4. Management, logging and team skills
If there are multiple locations, the cost and capabilities of the central management and central logging solution should be evaluated. Which platform your team knows better directly affects the risk of misconfiguration; even the most powerful device is left unprotected by faulty rules.
Ask a few practical questions during evaluation: can rule changes be rolled back, does the change history show who changed what, can the configuration be backed up automatically via CLI or API, does the log search tool return results quickly during an incident? On paper, all of these features exist. In the field, ease of use makes the difference. The team should be able to work on a platform not quickly but correctly; an administrator getting lost in the menus during a midnight outage means a faulty rule going live.
5. Support lifetime and security track record
- Check the vendor’s end-of-support timeline for the selected model.
- Look at how the vendor publishes security advisories and how quickly patches arrive.
- Verify that the management interface can be closed to the internet and supports MFA.
A model still in the sales channel may be one whose end-of-sale date has already been announced on the vendor’s product lifecycle page. If support for a device you plan to use for five years ends after three, the total cost calculation changes too; verify the dates on the vendor’s current lifecycle page. In the security track record, look not at whether vulnerabilities exist but at how the vendor behaves: are advisories published openly and on time, are workarounds provided, do patches arrive for more than one release branch?
Common mistakes in the selection process
- Buying a like-for-like replacement for the current device: The input for sizing isn’t the old device’s model but the features to be enabled today and in the coming years.
- Sizing only for the internet line: When internal segmentation, VPN tunnels and SSL inspection are enabled later, the device isn’t enough.
- Leaving HA for later: The second device needs to be the same model and the same software version; finding a match becomes harder after the model is discontinued.
- Not planning the migration at the quotation stage: How the old rule base will be moved to the new platform, which rules will be cleaned up and the outage window are part of the selection decision.
Rule migration tools speed up the work but also carry over faulty rules. Migration is the best moment to clean up the “all/any” rules and unused NAT definitions that have piled up over the years. Don’t carry the old baggage to the new device.
Short checklist for comparison
- Is the capacity measured with security profiles on at least 30–50% above the need?
- If SSL inspection will be used, is the capacity in this scenario sufficient?
- Has the 3–5 year total cost been calculated?
- Are HA, central management and logging needs included in the price?
- Will the team be able to manage the platform, and is training needed?
Eretek works vendor-neutrally on Fortinet, Palo Alto Networks, Cisco, Check Point, Sophos and other common platforms. For needs analysis and migration planning, see our Firewall Installation and Rule Management page, and for a Zero Trust roadmap, our Zero Trust Security Architecture page.