Cloud Services in Barcelona

Tourism and Hospitality / Hotel Technology

A Barcelona-based hotel technology company (SaaS platform serving 850 hotel properties across Spain, Portugal, France, and Italy — EUR 28M ARR, processing 12 million bookings per year through its platform) needed to modernise its cloud infrastructure to handle growth, reduce costs, and improve performance for hotel customers during peak booking periods. (1) Peak performance: the platform experienced extreme demand variation — booking searches peaked at 15x baseline during major events (MWC week, summer holiday booking season) and European holiday promotional periods. The existing single-region AWS deployment could not scale fast enough — peak period response times exceeded 3 seconds (from 300ms baseline), causing booking abandonment and hotel customer complaints. 8 incidents in the past year where the platform degraded significantly during peak periods — each incident affecting booking conversion and hotel revenue. (2) Multi-country latency: hotels in Spain experienced 150ms response times, but properties in France and Italy experienced 350-500ms due to all infrastructure being in a single AWS region (eu-west-1 Ireland). Italian hotel customers were particularly vocal about performance — competing platforms with Italian cloud presence offered better performance. (3) Cost growth: cloud spending had grown from EUR 1.2M to EUR 3.8M annually over 3 years — a 3.2x increase while revenue only grew 2.1x. The engineering team had over-provisioned infrastructure to handle peak periods (running peak capacity 24/7 rather than scaling dynamically), accumulated unused resources from failed experiments, and had no cost visibility by customer or product feature. (4) Data compliance: RGPD compliance was managed reactively — guest personal data was not classified or tracked across the cloud environment, data retention was inconsistent (some data kept indefinitely, other data deleted prematurely), and the platform processed guest data from 4 EU countries without country-specific compliance controls. The AEPD had recently increased enforcement of tourism data processing — creating regulatory risk. (5) Disaster recovery: the platform had no formal DR — the single-region deployment meant that an AWS eu-west-1 regional failure would take down the entire platform for all 850 hotel customers. Hotels would lose booking capability, guest check-in systems, and revenue management — with no estimated recovery time.

We redesigned and implemented the cloud platform over 12 weeks. (1) Multi-region architecture: three-region deployment — AWS eu-south-2 (Spain — primary for Spanish and Portuguese hotels), eu-west-3 (Paris — serving French hotels), and eu-south-1 (Milan — serving Italian hotels). Geographic routing: hotel traffic routed to the nearest region — reducing latency for all markets. Active-active: all three regions serving production traffic simultaneously — no single point of failure. Database: Aurora Global Database with read replicas in each region — write operations in the primary region (Spain), reads served locally. Conflict resolution: designed for the rare cases where cross-region writes occurred simultaneously. (2) Auto-scaling architecture: predictive scaling — ML model trained on 3 years of booking pattern data predicting demand 24 hours ahead. Pre-scaling: infrastructure scaled up before predicted demand peaks (rather than reacting to load). Spot instances: non-critical workloads (reporting, analytics, background processing) running on spot instances with automatic fallback to on-demand — reducing costs by 55% for these workloads. Serverless: event-driven functions for booking notifications, email processing, and webhook delivery — paying only for actual execution rather than provisioned capacity. Container orchestration: EKS-based microservices with horizontal pod autoscaling — individual services scaled independently based on their specific demand patterns. (3) Cost optimisation: resource cleanup — identified and eliminated EUR 680K in annual waste (unused instances, unattached EBS volumes, oversized databases, and forgotten development environments). Reserved instances: 3-year reserved instances for baseline capacity — saving 45% compared to on-demand. Spot strategy: 55% savings for interruptible workloads. Cost allocation: comprehensive tagging — cloud costs allocated to customer cohorts (by property count and feature usage), product features, and environments. FinOps dashboard: real-time cost visibility for engineering and finance teams — enabling informed decisions about feature development and resource allocation. (4) RGPD compliance: data classification — all guest personal data identified and classified across the cloud environment. Data residency: guest data processed in the cloud region of the hotel's country — Spanish hotel guest data in eu-south-2, French in eu-west-3, Italian in eu-south-1. Retention management: automated data retention policies per RGPD requirements — guest data retained for the legally required period and automatically purged. DPIA: data protection impact assessments completed for all processing activities — documented and maintained per AEPD expectations. (5) DR and resilience: active-active multi-region — any single region failure automatically handled by the remaining regions. Cross-region failover: automated failover with under 30 seconds of degradation — hotel operations continued without interruption. Chaos testing: regular failure injection to validate resilience — region failures, database failures, and dependency outages simulated monthly.

Performance: peak period response times from 3+ seconds to under 400ms (even during 15x demand peaks). Zero performance incidents in 12 months (from 8 incidents the previous year). Italian hotel customer satisfaction improved from 2.8/5 to 4.5/5 for platform performance. Multi-region latency: Spanish hotels from 150ms to 80ms, French hotels from 350ms to 95ms, Italian hotels from 500ms to 90ms — consistent sub-100ms performance across all markets. Booking conversion: 12% improvement in booking conversion rate (faster platform = more completed bookings) — translating to EUR 45M additional booking revenue for hotel customers annually. Cloud costs: from EUR 3.8M to EUR 2.4M annually (37% reduction) — while handling 40% more traffic. Cost per booking processed: from EUR 0.32 to EUR 0.17 (47% improvement). Waste elimination: EUR 680K in immediate savings from unused resources. RGPD: full compliance across all 4 country markets — data residency enforced, retention managed, and DPIA documentation complete. Zero AEPD findings during a subsequent data protection review. Resilience: zero downtime in 12 months — multi-region active-active handling 3 infrastructure failures (2 partial, 1 simulated full region) without customer impact. Customer growth: platform performance improvements and multi-region presence enabled expansion — 120 new hotel customers in 12 months (from average 80 per year), particularly in France and Italy where improved performance attracted previously hesitant hotels. Annual value: EUR 8.5M (customer growth — 120 new hotel customers driving additional subscription revenue), EUR 4.2M (booking conversion — 12% improvement translating to platform value for hotel customers), EUR 1.4M (cost reduction — EUR 3.8M to EUR 2.4M in cloud spending), EUR 1.2M (churn reduction — platform performance improving hotel customer retention by 6%), EUR 0.8M (compliance — RGPD compliance reducing regulatory risk and enabling expansion into stricter markets) = EUR 16.1M. Development cost: EUR 240,000. Payback: under 6 days.