Queue control for high-demand ticketing
Access control under high concurrency.
I implemented a queue layer to stabilize ticket sales under peaks of 10,000+ concurrent users.
Role
Software Engineer
Type
Ticketing & portals
Context
Context
This work ran inside ticketing portals for professional football clubs during high-demand sales windows. Before the change, there was no queue system, so users entered the sales flow at the same time and the portal broke under load.
Problem
Problem and constraints
The immediate problem was uncontrolled concurrency: timeouts, failed purchases and lost revenue once traffic hit around 10,000 concurrent users.
The hard constraint was an external LALIGA platform sitting inside the purchase flow. It didn't scale under peak load and we couldn't modify it, so any viable solution had to throttle access before reaching it.
The problem was already hitting the business: a commercial account had been lost after exactly this kind of failure during a previous on-sale.
Approach
Approach and technical decisions
I evaluated queue providers and selected Queue-Fair based on cost, fairness in the waiting room, and evidence of use at higher scale.
I handled the vendor conversation and agreement directly in English, then implemented Queue-Fair through a frontend script on WordPress because it gave faster rollout, simpler maintenance, and direct operational control.
The implementation combined that script integration with a decoupled match landing page, so access control could be activated and adjusted from WordPress without backend changes.
Architecture
Architecture
- 1User
- 2WordPress landing
- 3Queue-Fair waiting room
- 4LALIGA ticketing platform
- 5Confirmation
Decisions
Key decisions
01
Shielding a non-scalable external dependency without blocking sales.
02
Operational integration that could be controlled directly from WordPress.
03
Reducing pressure on the core system during peak traffic.
Outcome
Outcome
The sales flow stopped collapsing at around 10,000 concurrent users and moved to an ordered waiting room of roughly 60,000 users.
That gave the ticketing operation a controlled entry point and supported larger commercial accounts afterwards.
The queue protects the external platform but doesn't make it faster — when LALIGA's side slows down, the queue's progress slows with it. We accepted that trade-off explicitly: ordered failure beats uncontrolled collapse.