What Is a Performance Enhancing Proxy (PEP)? How TCP Splitting Speeds Up Satellite and Inflight Internet
If you have ever read about satellite broadband, maritime VSAT or inflight Wi-Fi, you have probably come across the term PEP, short for performance enhancing proxy. It sounds technical, but the idea behind it is simple and the results can be dramatic.
This guide explains what a PEP is, how the technique known as TCP splitting works, where it helps, where it does not, and how to check whether it is making a difference on your own connection.
The problem a PEP is designed to solve
Most of what you do online, from loading web pages to sending email, travels over TCP. TCP is reliable because it is careful: the sender transmits some data, waits for the receiver to confirm it arrived, and only then sends more. It also starts slowly and ramps up gradually, a behaviour known as slow start.
On a home broadband line, each of those confirmation cycles takes a few milliseconds, so you never notice. On a geostationary satellite link, a single round trip commonly takes more than half a second. Multiply that by the handshakes, acknowledgements and ramp-up steps involved in loading one page, and the delays pile up.
The frustrating part is that the link may have plenty of capacity. TCP simply never gets the chance to use it, because it spends most of its time waiting for replies that are still travelling through space.
How a PEP works: TCP splitting in plain English
A performance enhancing proxy sits in the path of your traffic and changes how TCP behaves across the slow part of the journey. The most common technique is TCP splitting.
Instead of one long TCP connection stretching from your device all the way to a website, the PEP breaks it into separate segments:
- A short, fast segment between the PEP and the website, where round-trip times are low and TCP works normally.
- A long-delay segment across the satellite or high-latency link, where the PEP can use settings and congestion control methods tuned for long round trips.
- A local segment at your end, where acknowledgements can be answered quickly rather than waiting for the far side.
Because each segment is managed separately, the website sees a fast, responsive client, and data can be pushed across the slow link far more aggressively than standard TCP would dare. Specialised congestion control algorithms, such as TCP Hybla, were developed precisely for this kind of long-delay path.
One-ended and two-ended PEPs
Some PEPs work from a single point in the network. Others are distributed, with processing at both ends of the slow link so they can optimise traffic in both directions. Two-ended designs generally have more control, but need equipment or software on each side.
What a PEP can and cannot fix
It is worth being realistic. A PEP is very good at some things and has no effect on others.
| Situation | Does a PEP help? |
|---|---|
| Web pages loading slowly on a high-latency link | Yes, often significantly, because page loads involve many TCP round trips |
| Downloads that never reach the link's full speed | Yes, by keeping more data in flight across the long-delay segment |
| Email clients syncing sluggishly | Usually, as email protocols are chatty and sensitive to round trips |
| Voice and video calls feeling laggy | Only partly. The physical distance to a satellite still adds delay that no proxy can remove |
| A link that is simply full | Not on its own. You also need to reduce the amount of traffic competing for it |
That last row matters. On shared, costly links, a surprising share of traffic is background data nobody asked for: adverts, trackers, telemetry and malicious scripts. Removing that junk frees capacity in a way that TCP tuning alone cannot.
The catch: encryption and VPNs
Classic PEPs need to see and manipulate TCP traffic. That raises two well-known complications.
- VPN tunnels. When you use an ordinary VPN, your TCP connections are wrapped inside an encrypted tunnel. A PEP run by the network provider usually cannot see inside, so it cannot split those connections, and performance can drop noticeably.
- Newer protocols. Some modern traffic, such as QUIC, runs over UDP and encrypts its transport details, so traditional TCP-splitting proxies cannot optimise it in the same way.
This is why simply switching on a standard VPN on a satellite link can make things feel slower, and why it helps when acceleration is built into the service you connect to rather than left to whichever network you happen to be on.
How FastAPN uses PEP acceleration
FastAPN, developed by Transcom and Flytlink, is built around this idea. It is a PEPsal VPN: a performance enhanced proxy for high-latency and congested networks, combined with a Deep Packet Filter (DPF) that destroys unwanted data at source before it travels to you.
In practice that means:
- TCP-split acceleration and real-time optimisation aimed at satellite (GEO, LEO, MEO, BGAN and SBB), aviation (ATG, EAN and SATCOM) and any other high round-trip-time network. FastAPN quotes up to 95% latency reduction on GEO satellite links and 10 to 15 times speed gains on high-RTT connections.
- Filtering of adverts, tracking, telemetry, malware and phishing domains upstream, so less of your share of the link is wasted.
- All protocols, ports and services are enhanced, including email, Exchange, Outlook and streaming.
- No hardware or software changes. You add the connection details to your device's built-in VPN settings. The server is pep.fastapn.com, using L2TP/IPsec, and it works with phones, tablets, laptops, routers and embedded systems.
- Dynamic nodes. The service uses the access node closest to you and switches dynamically. On a terrestrial connection it behaves as a normal VPN with the DPF still active.
For airlines and operators, FastAPN's inflight server options include a compact Remote Service Node that travels with the aircraft or user, giving packet processing on both sides of the link, much like the two-ended designs described above. Fleets needing more than three devices can look at corporate and block pricing.
How to tell whether it is helping
The best test is your own connection. A sensible approach:
- Load a few everyday pages and note how long they take without any acceleration.
- Connect using the free access details on the configuration guide. The free connection resets every three minutes, which is enough for a quick comparison.
- Once connected, use the live dashboard at pep.fastapn.com to see your upload and download speeds and connection status.
- Run the unwanted data check. A score below 80% suggests your connection is being flooded with adverts, tracking and other junk.
Repeat the same pages and compare. On a low-latency home line you may notice little beyond the filtering; on a satellite or inflight link the difference should be much clearer.
Choosing a FastAPN plan
If the free test shows a benefit, FastAPN offers a 24 Hour Pass for £2 (a one-off payment with no auto-renewal, ideal for a single flight), a Monthly plan at £7.49 that you can cancel anytime, and an Annual plan at £59. All include unlimited bandwidth across all network types, and credentials are emailed to you after purchase.
You can compare the options and get connected on the FastAPN plans and pricing page.