You have a 1 Gbps fibre line and yet, when someone uploads a big file from the office, the client video call freezes or the CRM takes forever. It's not that fibre is slow. It is bufferbloat, the most common (and least understood) problem of the modern Internet.
What bufferbloat actually is
Routers, modems and network gear have buffers — temporary memory for queueing packets that arrive faster than they can be forwarded. For decades, vendors made these buffers larger and larger, believing it would avoid packet drops.
The flaw: big buffers + standard TCP = catastrophic latency under load. When the link is saturated by a download, voice and browsing packets queue behind thousands of download packets and take 500-2000 ms to get out. The connection feels slow even though raw throughput is high.
How it's measured: the DSLReports test and the A+ grade
The standard test is the Bufferbloat Test (dslreports.com/speedtest or Waveform bufferbloat). It uploads and downloads simultaneously while measuring added latency. Results are graded with a letter:
- A+: < 5 ms added latency (excellent)
- A: < 30 ms
- B / C: 30-200 ms (noticeable)
- D / F: > 200 ms (unusable for real-time voice/video)
Most Spanish operators score excellent when idle, but drop to B or C under real load. The 100-300 ms hit kills perceived quality.
The fix: AQM (Active Queue Management)
The answer is not smaller buffers — it is smart queue management. Modern AQM algorithms (CoDel, FQ-CoDel, PIE, L4S) decide which packets to drop and which to keep based on how long they've sat in the queue and which flow they belong to.
FQ-CoDel (Fair Queueing with Controlled Delay) is the de-facto standard. It gives a separate queue per flow, so a download can't interfere with a video call. Combined with CoDel, it controls latency by intelligently dropping packets when the queue grows.
What Olivet FlowEngine does
Olivet FlowEngine is our proprietary TCP optimisation engine — a transparent TCP proxy (RFC 3135) running on the backbone. Among other things, it implements FQ-CoDel tuned for the business traffic profile and cooperates with L4S / TCP ECN to work even better with modern endpoints.
Measurable result on our connections: bufferbloat from B/C to a consistent A+, under-load latency < 5 ms, TCP retransmissions down by 80%. But don't take our word for it: run the test at waveform.com/tools/bufferbloat on your current line and compare against the Olivet line once installed.
Conclusion
Bufferbloat is the difference between "I have fast Internet" and "I have Internet that actually works well". A 1 Gbps fibre without modern AQM is like a 500 HP car running summer tyres on icy roads: powerful on paper, disastrous in practice.
Frequently asked questions
Can I enable FQ-CoDel on my current router?
Routers running OpenWrt can; most operator-supplied routers can't, or only partially. Olivet FlowEngine runs on the backbone, so it works regardless of the customer router.
Does bufferbloat only affect homes with lots of devices?
No. One HD video call plus a user downloading updates is enough to trigger bufferbloat on residential lines without AQM.
Want this quality at your business?
1 Gbps symmetric fibre certified to Y.1564. €149/month. No lock-in.
More articles
What is Y.1564 and why it changes business fibre
Y.1564 is the ITU-T standard that certifies speed, latency, jitter and loss of your fibre. We cover the 5 KPIs and why yours should ship with one.
ORO and QoS 802.1p: how a business gets priority over the neighbour
The ORO profile on Telefónica's NEBA network gives businesses QoS p-bit 3 priority over residential Best Effort. How it works, technically.
99.999% SLA: what it really means (and how it's kept)
99.999% uptime equals 5 min 16 s of downtime per year. How it's measured, how HA redundancy delivers it, and what your SLA should say.