A firewall’s headline throughput is a useful first filter, not a safe capacity plan. The number may describe basic packet forwarding under a specific test profile. Your production requirement may include intrusion prevention, application control, malware inspection, TLS decryption, site-to-site VPN, remote access, or all of them at once. Size the device against the traffic and protections you actually intend to run.
Why one throughput number can mislead
Firewall data sheets often publish several performance figures because they describe different packet sizes, protocols, and enabled services. A stateful firewall test may not represent a policy that also performs intrusion prevention and application inspection. TLS inspection adds another workload. VPN throughput is a separate metric with its own test conditions.
For example, Fortinet’s May 2026 Product Matrix lists these figures for the FortiGate 90G:
| Published metric | FortiGate 90G figure | What to check |
|---|---|---|
| Firewall throughput (1518-byte UDP) | 28 Gbps | Basic forwarding test profile and packet size |
| NGFW throughput (Enterprise Mix) | 2.5 Gbps | Feature bundle and traffic mix defined by the vendor |
| Threat Protection throughput (Enterprise Mix) | 2.2 Gbps | Which threat services are included and enabled |
These figures are not interchangeable or a promise of your application’s real-world speed. They illustrate why the metric label, test method, software release, and enabled features matter. Cisco likewise notes that performance varies with enabled features, protocol mix, and packet size. Always read the footnotes and compare like with like.
Build a workload profile before comparing models
Start with the traffic that must cross the firewall in both directions. Record current busy-hour throughput, expected growth, internet circuit speeds, inter-site traffic, and whether traffic is centralized or split across links. If the requirement is based only on the WAN circuit’s advertised rate, verify that it reflects measured peak use and the traffic routed through the appliance.
- Inspection policy: Identify whether IPS, application control, antivirus or malware inspection, web filtering, and TLS decryption are required. Ask for published performance under the closest matching feature set.
- Encrypted traffic: Estimate the share of traffic that will be decrypted and inspected, and specify the protocols, certificate handling, exclusions, and privacy requirements. Do not assume a generic “threat protection” number includes your intended TLS policy.
- VPN: List site-to-site tunnels, expected remote users, and the bandwidth that needs protection. Check VPN throughput, peer or user limits, and the test conditions separately.
- Connections: Estimate concurrent sessions and new connections per second, especially for environments with many short-lived sessions, guest devices, or internet-facing services. Peak bandwidth alone will not describe these limits.
- Interfaces and layout: Confirm port count, media type, link speeds, management ports, VLAN or routed-interface needs, and whether transceivers are included or separately sourced.
Fortinet’s FortiOS sizing guidance explicitly asks administrators to consider users, devices, interfaces, threat-protection throughput, and how deep SSL/SSH inspection can change performance. Treat the sizing guide and the exact model data sheet as a pair: one frames the workload, the other supplies test figures.
Account for the failure case and the next upgrade
For a high-availability pair, size around the capacity available after the failure mode you need to survive. Do not assume two devices automatically provide twice the protected throughput, or that every session survives a failover. Confirm the vendor’s supported HA design, licensing, synchronization behavior, and the performance of a single remaining unit with the intended inspection profile.
Include a realistic growth target and planned circuit changes in the requirements. If traffic is expected to move from 1 Gbps to 2 Gbps, for example, have the vendor quote and substantiate the required inspected throughput for that future state. Avoid relying on an arbitrary oversizing percentage when a documented growth horizon is available.
Use a like-for-like validation step
Before purchase, ask the manufacturer or supplier to map your required services to the correct data-sheet metrics and current software version. For a short list, arrange a proof of concept or a documented reference design where practical. Define the traffic mix, inspection settings, VPN load, and acceptable latency before testing; otherwise, a result may not answer the sizing question.
Quick RFQ checklist: device count and user profile; measured busy-hour traffic; current and planned WAN speeds; security services and TLS-inspection scope; VPN tunnels and remote users; session and connection-rate needs; interface and transceiver requirements; HA failure target; software and license term; and the performance figure required with those features enabled.
Turn the figures into a defensible shortlist
Compare firewalls only after the requirement is clear. Match the same protection features, traffic assumptions, VPN use, and failure target across candidates. Then check licensing, interface fit, support terms, and lifecycle requirements alongside performance. A spec sheet can narrow the field; it cannot replace a workload profile or a validation plan.
For a broader firewall range, see our firewall selection. For a security architecture reference, read Cisco Security Reference Architecture. You can send a parts list or requirements through our enterprise procurement page, or contact our team to discuss a quote.
Sources
- Fortinet Product Matrix (May 2026) — model performance metrics and footnotes.
- Fortinet FortiOS 7.4.4 NGFW ATP Deployment Guide: Sizing — workload factors and inspection considerations.
- Cisco Secure Firewall 200 Series Data Sheet — feature-specific throughput and performance caveats.
- Cisco Newsroom photo source and Cisco Newsroom reuse terms.


















