Limits: Using pfSense Traffic Shaper to Cut Costs on Low‑Traffic Workloads

Direct Answer

Enabling the pfSense Traffic Shaper will not let you downgrade to a cheaper ISP plan or a lower‑power NIC in typical low‑traffic deployments. The shaper only enforces egress limits; it does not change the provisioned line rate your ISP bills for, nor does it trigger NIC power‑saving states unless the hardware and driver explicitly support Energy Efficient Ethernet (EEE) and the link sits idle for extended periods.

Confirmed Facts

• CPU overhead is negligible at low PPS. ALTQ/HFSC classification cost scales with packets per second, not the configured bandwidth ceiling. At <1 Mbps average (≈1–2 kpps) modern x86_64 CPUs see <1–2% extra interrupt/softirq time.
• Shaper latency is controllable but fragile. HFSC real‑time curves can guarantee sub‑millisecond latency for VoIP/DNS, yet mis‑tuned linkshare or upperlimit curves often add 1–5 ms jitter under burst.
• NIC power draw stays at line rate. Most server NICs (Intel i350, i210, Mellanox ConnectX) maintain 1 Gbps/10 Gbps link speed regardless of shaper‑enforced throughput. EEE or ASPM must be supported by the NIC, driver, and switch partner — and the link must be near‑idle for minutes to enter low‑power states.
• ISP billing models dictate savings. If your ISP bills on 95th‑percentile or committed information rate (CIR), a hard shaper cap avoids overage fees but does not lower the base charge. Only plans billed purely on total GB transferred can benefit from a sustained cap.

Likely Explanations (Unverified)

• Aggressive shaping (e.g., 10 Mbps limit on a 1 Gbps port) may reduce switch‑port power slightly if the switch honors EEE, but this is vendor‑specific and rarely measurable.
• ALTQ is deprecated in FreeBSD 14+/pfSense 2.7+; the AQM limiters (fq_codel, cake) offer lower overhead and better latency control, but they also do not affect NIC link speed or ISP provisioning.

Verification Steps (Run on Your Hardware)

1. Confirm queue config and counters:

   pfctl -s queue -v

2. Measure CPU impact at realistic PPS:

   # Terminal 1: generate load
   iperf3 -c  -t 60 -P 4
   # Terminal 2: watch CPU breakdown
   systat -vmstat 1

   Compare interrupt/softirq % with shaper enabled vs. disabled.
3. Measure latency under load:

   flent rrul_be -l 60 -H 
   # or quick ping test
   ping -i 0.01 -c 1000 

4. Check NIC EEE support:

   pciconf -lv | grep -A5 -i ethernet
   sysctl dev...eee
   ifconfig  media

One Missing Diagnostic Detail

What is your ISP's exact billing model (95th‑percentile, CIR, flat‑rate with overage, or pure GB‑metered)? This single fact determines whether a shaper‑enforced cap can reduce your monthly bill. If the plan is CIR or 95th‑percentile, the shaper cannot lower the base cost; if it is pure GB‑metered with no minimum commit, a sustained cap may help.