Continuous Latency Verification: Why Weekly Compliance Testing Doesn't Scale (yet)

Part 1 of 2: The Fairness Attestation Gap by Eric Dorre, co-Founder - FairConnect

Every colocation agreement makes the same implicit promise: equal access, equal latency, equal treatment for every rack in the room. The most sophisticated trading firms don’t take that promise on faith. They don’t just accept whatever equidistant cross-connect they’re assigned and call it done. They verify their own line continuously, running timing probes, checking timestamps and "shooting their lasers back" at the exchange to flag anything out of tolerance. Some go further, ordering more cross-connects than they need, keeping whichever perform best, and cancelling the rest, not only when conditions shift but as a matter of course. They assume they are at a disadvantage to someone else, and they do all in their control to mitigate that risk.

With highly deterministic exchange infrastructures, that edge is stable enough to be worth chasing. It’s exactly the right instinct, and one I was maniacal in pursuing when I was on that side of the table. But it only answers one question: am I within tolerance? It can’t answer the question that defines fairness: is everyone else? No client can see their peers’ numbers. A firm can spend heavily to continuously optimize for the shortest possible cross-connect, and it still comes down to trust: trust that no rival with comparable alpha simply got a luckier cable run, and trust that no rival is quietly out of tolerance, or cheating outright. Most exchanges cannot credibly settle that question at scale either. This series will explain why this so difficult, and give exchanges, their liquidity providers and regulators promise that the firm I co-founded, FairConnect, can close the expectation gap.

This isn’t just an expectation clients have quietly formed. It’s a standing regulatory obligation, and a more precise one than the opening line lets on. Under EU MiFID II’s colocation rules (Commission Delegated Regulation 2017/573, adopted post-Brexit in the UK), EU and UK trading venues must "take all reasonable steps to monitor all connections and latency measurements" and demonstrate non-discriminatory treatment among clients on the same service tier. In the US, the same principle is built into each exchange’s own SEC-filed colocation rules under the Exchange Act’s fair-access standard. Regulators are actively widening their focus here, not narrowing it: the FCA opened a consultation in September 2025 over whether a major venue’s datacenter access was being fairly administered from the rooftop to client racks. The expectation already exists on paper. What hasn’t caught up is the standard of proof: policies and procedures, backed affirmatively by auditable evidence, that the exchange is monitoring and maintaining latency fairness. What’s missing for most exchanges is a way to meet that obligation continuously and at scale.

Picture a mid-size exchange with a few hundred cross-connects carrying order entry and market data to co-located clients. Verifying every one of those paths still delivers equal latency (within a stated SLA), using industry-typical manual measurement, is not a quick walkthrough. It runs into the tens of person-hours per cycle: an engineer with a handheld instrument, working connector by connector, patch panel by patch panel. Put rough numbers on it: five minutes per connector for setup, measurement, and logging means three hundred cross-connects takes roughly twenty-five person-hours before anyone reconciles the readings or chases an outlier. Grow the colocation floor, as most venues plan to, and the labor grows with it in a straight line or worse. None of that hour count yet includes the follow-up any careful engineer runs when a reading looks borderline: a second instrument, a second technician, a second look before signing an attestation that could end up in front of a regulator, not to mention broken connectors (or the periodic break). And anyone who's ever worked in or managed a data center knows the part nobody puts in the SLA: humans doing repetitive, manual work at scale make mistakes. A mislabeled port, a transposed reading, a skipped row in the log, not from negligence, just from what happens when a process this repetitive depends on people instead of machines. Run that cycle weekly or even monthly, as most attestation frameworks assume, and the arithmetic stops cooperating with the calendar. "Weekly" quietly becomes "whenever the team can get to it."

This is not a story about anyone cutting corners. It's a story about an attestation model built for periodic spot-checks trying to keep pace with infrastructure that trades continuously, in an environment where nanoseconds decide who trades. The gap between what regulators and clients expect and what manual verification can realistically deliver is structural, not a failure of diligence. It's the gap any industry hits when a process built for occasional checks meets a system that never stops running. As markets go 23x5 and eventually 24x7, the impractical becomes the impossible.

None of this needs to be a defensive story, for the venue that gets ahead of it. An exchange that moves first to continuous, automated, always-on verification isn't just closing a compliance gap. It's converting an obligation that trading firms and regulators currently take on faith into something they can see and trust in real time. In a market structure moment this fluid, the shift from "blind faith" to "we’ve got the data and audit trail to demonstrate fairness” is a genuine point of competitive differentiation, not a cost center.

The open question this raises, and the one we'll take on in Part 2: even when a venue does test continuously, how does it know the measurement itself is trustworthy? The instruments doing the measuring don't always agree with each other, and that disagreement can complicate matters more than most infrastructure teams realize.

If you are an exchange operator and your last full cross-connect latency verification cycle took longer than a week to complete, you already know which side of this gap you're on. Similarly, for Ultra Low Latency shared network providers who may have no regulatory mandate: you all share the implicit promise to treat clients the same from the exchange or meet-me-room to the client rack. We'd welcome the conversation with FairConnect to support your efforts in monitoring and demonstrating your delivery on that fairness promise.