Fewer support tickets
Many "Internet is slow" complaints aren't about bandwidth: they're about sporadic losses misread by TCP. FlowEngine absorbs them before they become a ticket.
Home›Olivet FlowEngine
Proprietary TCP optimisation engine
Every bit, at peak performance
We get peak performance out of every bit of your connection. The Internet is not perfect: remote servers, intermediate congestion and even your own WiFi introduce imperfections that standard TCP misreads. Olivet FlowEngine is the proprietary engine we built to extract real value from every Mbps of your fibre.
Why it matters
Three tangible benefits for your business
We don't talk about technology for its own sake. This shows up every day at the office.
Video calls and SaaS without drops
Interactive applications (Teams, Meet, VoIP, cloud ERP) get automatic priority. Under load, your office keeps flowing.
More out of your contracted Gbps
Without changing a single cable in your office, FlowEngine speeds up page loads by 19-42% and removes the bufferbloat that made your 1 Gbps feel slower.
Measurable results
Four metrics, all auditable
Data can be reproduced with any public tool: speedtest, waveform.com/tools/bufferbloat, WebPageTest.
−50/83%Latency
−80%Retransmissions
+19/42%Page loads
B → A+Bufferbloat
How it works
Four parts, one engine
Each FlowEngine module solves a specific real-network problem.
A · Transparent TCP proxy
Integrated in the network, invisible to the client
FlowEngine runs inside Olivet Telecom's infrastructure, between your office and the wider Internet. Per RFC 3135, it terminates and re-opens each TCP session applying our optimisations without changing anything on the client: no agent, no configuration, no compromise on end-to-end encryption (TLS).
- Zero config on PC, server and mobile
- Native IPv4 and IPv6 support
- No content inspection for encrypted traffic — acts at transport layer
- Can be disabled on request for A/B testing
B · Real congestion vs. WiFi loss
The algorithm that knows the difference
Standard TCP (CUBIC) reacts the same way to any packet loss: aggressively slowing down. The problem? In your office, most losses come from WiFi (interference, noise), not from real congestion. Collapsing speed is the wrong response.
FlowEngine analyses transmission-rate trends in short windows and distinguishes between the two cases. When loss is accidental, it holds throughput. When it's real congestion, it backs off with precision.
- Real-time classifier of loss vs. congestion
- Fast recovery without window collapse
- ECN support for explicit signalling
- Auto-tuning to the line's profile (RTT, BDP)
C · Per-application priority
Video calls first, backups after
Not all traffic is equal. A video call or a VoIP session tolerates very little latency but little bandwidth; a backup or a Docker pull tolerates latency but demands all the bandwidth it can get. FlowEngine recognises these patterns and treats them differently.
It does so without inspecting content: packet size, cadence, RTT and window sizes are analysed to classify each flow. Patterns are kept up to date from our NOC.
- Five classes: real-time, interactive, streaming, bulk, background
- Dynamic reallocation when flow type changes
- Soft guarantees for critical class under load
- No content DPI — patterns only
D · Smart queues (AQM)
Goodbye bufferbloat
Bufferbloat is the hidden cause of "slow Internet under load": critical packets wait behind huge buffers full of bulk traffic. FlowEngine implements enhanced FQ-CoDel with one queue per flow and L4S support (Low Latency, Low Loss, Scalable throughput), keeping latency under load below 10 ms.
- One queue per flow — a slow flow never blocks the others
- CoDel AQM with adaptive thresholds
- L4S and ECN for apps that support them
- Result: bufferbloat from B to A+ on standard tests
Comparison
Standard TCP vs. Olivet FlowEngine
Details matter when your whole office depends on it.
| Behaviour | Standard TCP (CUBIC) | Olivet FlowEngine |
|---|---|---|
| Loss detection | Always slows down | ✓ Loss vs. congestion aware |
| Response to WiFi jitter | Throughput collapse | ✓ Holds speed |
| Per-application priority | None — all traffic equal | ✓ Five automatic classes |
| Bufferbloat | Single queue, latency >150 ms | ✓ FQ-CoDel + L4S, <10 ms |
| ECN / L4S support | Limited / disabled | ✓ On by default |
| Client configuration | Not applicable | ✓ None — 100% transparent |
| Gain certification | Self-reported | ✓ Y.1564 auditable |
Frequently asked
Technical questions
What systems teams usually ask us.
Do I need to change my router or configure anything on the client?
No. FlowEngine is a transparent TCP proxy (RFC 3135) integrated in our network. It works on any device and operating system without changing client configuration.
Does it work with IPv6 and encrypted traffic (HTTPS/TLS)?
Yes. FlowEngine operates at the TCP transport layer without opening or inspecting encrypted content. It supports native IPv4 and IPv6 and does not break the TLS chain.
Does it affect traffic privacy or neutrality?
No. FlowEngine optimises transport parameters (windows, ACK, queues, ECN) without decrypting traffic or altering content. Per-application classification relies on flow patterns, not on content.
Does it work over VPNs?
Yes for TCP-based VPN traffic (e.g. HTTPS tunnelling). For UDP-only VPNs (WireGuard, IKEv2) optimisation is reduced: FlowEngine applies AQM queue management only, not TCP window optimisation.
What overhead does it add to minimum latency?
Less than 1 ms under normal conditions. Under accidental loss, net effective-latency gain can reach −83% thanks to the elimination of unnecessary retransmissions.
How are gains validated on your network?
Every connection ships with an ITU-T Y.1564 certification report with auditable measurements of CIR, EIR, FTD (latency), FDV (jitter) and FLR (loss). FlowEngine gains can be contrasted by disabling the engine on request and repeating the test.
Try it
Prove it in your office
Every Olivet Telecom connection ships with FlowEngine enabled by default and an auditable Y.1564 report. No lock-in, no small print.