Skip to main content
All case studies
Finance
6 regions, 4 continents · 21 engineers

Apex Trade

Global VPS fleet, fail-over rehearsed every Friday.

Apex Trade routes orders for retail FX customers across six global regions. Their previous infra was a single hyperscaler region with a hot DR site they had never actually failed over to in three years. After a near-incident in late 2025, they rebuilt their stack on Vintony with multi-region active-active and a fail-over runbook they now rehearse weekly.

Pools

6

Active-active across 4 continents

p95 APAC latency

44ms

Was 240ms+

Fail-over drills

weekly

Rehearsed every Friday 10am UTC

Customer-impacting incidents

0

Since the rebuild

The problem

The pre-Vintony setup was a single-region Frankfurt deployment with a hot-standby in Ashburn. Latency to APAC customers was 240ms+; routing them through Frankfurt was bleeding conversion. The DR site existed on paper, but the runbook to fail over was untested in production. After a 47-minute Frankfurt outage in November 2025 that did not trigger automatic failover — because of a misconfigured health check that nobody had touched in two years — the leadership made multi-region active-active a Q1 priority.

The migration

Apex rebuilt as six regional VPS pools: Frankfurt, Ashburn, San Francisco, Singapore, Tokyo, Sydney. Each region runs identical stacks, sharing state via a globally-replicated Postgres with regional read replicas and a Redis cluster pinned to the customer's home region. Order flow routes by geo to the closest healthy pool; if a pool fails health checks for thirty seconds, traffic shifts. They run a synthetic fail-over rehearsal every Friday at 10am UTC by manually draining one pool. After six months of weekly drills the team has never been surprised by a real incident.

We needed somewhere we could rehearse a regional failure without theatre. Vintony's per-region isolation makes the drill cheap enough that we actually do it.

Imani Okafor, Principal SRE, Apex Trade

What's next

Apex are evaluating a São Paulo pool for their growing LATAM customer base and planning a chaos-engineering programme that introduces correlated failures across two regions — something their previous infra would never have survived.

Could we tell a story like this for you?

More stories