Pete Ratkevich Case Study: Media Server
Cloud Migration & Global Content Delivery for 200M+ Digital Assets
Background
The Media Server was the central operational circulatory system for Pearson's higher education digital products, storing, processing, and serving high-resolution images, video lectures, interactive animations, and audio narrations to millions of students worldwide. Over a decade of digital courseware growth had accumulated more than 200 million digital assets totaling 30 TB of production data.
The underlying infrastructure, however, was anchored to aging on-premises physical servers and NetApp SAN/NAS storage clusters. As higher-ed enrollments transitioned to digital-first learning, seasonal back-to-school traffic surges pushed physical disk I/O to capacity. High latency in international regions degraded student learning experiences, hardware refresh cycles consumed capital budgets, and single-datacenter dependencies created catastrophic disaster recovery risks.
To eliminate these constraints, Pearson initiated an enterprise migration to Amazon Web Services (AWS). As Product Manager, I led the product strategy, cross-functional execution, risk mitigation framework, and stakeholder coordination to migrate 200M+ production assets to the cloud, achieving zero downtime and maintaining seamless continuity across ongoing editorial publishing pipelines.
Project Overview
The primary objective was transforming a fragile on-premises storage environment into an elastic, globally distributed digital asset backbone. The scope encompassed migrating petabyte-scale media repositories, modernizing ingestion APIs, and establishing real-time cloud observability without interrupting in-flight courseware production:
- Cloud Object Storage Architecture: Engineered Amazon S3 storage tiers with automated lifecycle policies to balance high-speed access with cost-efficient archival.
- Worldwide Edge Distribution: Deployed Amazon CloudFront CDN edge caching to terminate latency bottlenecks for international students and instructors.
- Zero-Downtime Dual-Run Proxy: Architected an active-active ingestion layer that synchronized assets across on-premises and AWS environments with continuous MD5 checksum verification.
- Infrastructure as Code & CI/CD: Implemented automated deployment pipelines using Docker, Jenkins, and AWS Lambda event triggers for hands-free media optimization.
- Proactive Telemetry & Observability: Integrated New Relic APM, Grafana, and AWS CloudWatch dashboards for real-time monitoring of throughput, cache-hit ratios, and error budgets.
Challenges Faced
1. Data Scale & Bit-Level Integrity:
Migrating 200 million heterogeneous assets (including high-resolution scientific diagrams, audio clips, and interactive Flash/HTML5 packages) across 30 TB represented an enormous data-integrity risk. A single corrupted diagram or dropped audio file could render an online exam question unanswerable in an active college course. We required automated verification at the byte level.
2. Uncompromising Business Continuity:
Pearson's editorial, authoring, and vendor teams operate around the clock on tight semester publication deadlines. Taking the Media Server offline for a traditional multi-day cutover window was completely off the table. The migration had to occur invisibly in the background while thousands of active titles were being edited.
3. Legacy Upstream Dependencies:
Dozens of internal authoring tools, assessment engines, and LMS packaging systems had hardcoded dependencies on physical NetApp directory paths and proprietary storage protocols. Modernizing the backend required building backwards-compatible API adapters so upstream systems could continue operating without requiring coordinated engineering rewrites.
4. Severe Seasonal Concurrency Surges:
During the first two weeks of the fall and spring semesters, millions of students simultaneously log in, launch courseware, and stream media assets. The new cloud architecture needed to handle extreme traffic spikes elastically without manual provisioning or server degradation.
Solutions and Execution
1. Architecture Modernization Comparison:
We replaced rigid, single-point-of-failure hardware with an elastic, multi-region cloud architecture designed for high availability and low latency:
| Architecture Dimension | Legacy On-Premises (NetApp) | AWS Cloud Platform | Operational & Business Impact |
|---|---|---|---|
| Storage Elasticity | Physical drive shelves; multi-month hardware procurement and manual SAN provisioning. | Amazon S3 elastic storage tiers with intelligent lifecycle rules. | Eliminated capital expenditures and removed storage capacity caps entirely. |
| Global Content Delivery | Centralized datacenter origin; high cross-regional latency and bandwidth saturation. | Amazon CloudFront CDN with edge locations distributed globally. | Slashed international media load times by over 65% across web and mobile. |
| Reliability & Disaster Recovery | Single-datacenter failure domain; scheduled tape backups with multi-day RTO. | Multi-AZ redundancy with 99.999999999% (11 9's) data durability. | Zero risk of asset loss; total resilience against regional datacenter outages. |
| Ingestion & Media Processing | Monolithic servers handling ingestion, virus scanning, and asset transcoding synchronously. | Serverless AWS Lambda microservices triggered asynchronously on S3 upload events. | Decoupled ingestion pipelines; eliminated file processing bottlenecks during rush windows. |
2. Active-Active Dual-Run Ingestion Pipeline:
To eliminate the risk of cutover failure, we established a phased, active-active dual-run strategy. When content producers uploaded an asset, custom middleware routed the payload simultaneously to the legacy NetApp array and Amazon S3.
Background worker processes continuously executed asynchronous MD5 checksum hashing against both storage endpoints to ensure 100% bit-level parity before redirecting downstream read traffic. This dual-run approach operated for 90 days, enabling comprehensive testing under production workloads without risking data loss or workflow interruption.
3. Automated Deployment & Observability:
We codified our deployment pipelines using Docker containers and Jenkins CI/CD automation, ensuring that configuration environments remained identical across staging and production. To ensure total operational visibility, we configured real-time telemetry:
- Amazon CloudWatch & Grafana: Real-time dashboard tracking edge request rates, S3 latency percentiles, and origin shield offload efficiency.
- New Relic APM: Transaction profiling isolating slow API queries and monitoring database connection pools on Amazon RDS.
- Automated Incident Response: Configured alerting thresholds that automatically notify on-call engineers of anomalous 4xx/5xx spikes before users experience degradation.
Outcomes and Business Impact
The Media Server migration delivered immediate performance gains and transformed digital media operations across Pearson:
- Zero Asset Loss Across 200M+ Files: Achieved 100% bit-level verification across 30 TB of production assets, completing the entire migration with zero lost files or corrupted media.
- Zero Production Downtime: Maintained continuous 24/7 availability for active editorial workflows throughout the 90-day dual-run and final cutover.
- 65% Reduction in Global Media Latency: Amazon CloudFront edge caching delivered sub-1.2 second average media load times worldwide, vastly improving courseware responsiveness in remote and international regions.
- Transformed Cost Structure: Replaced unpredictable, multi-million-dollar physical hardware refresh cycles and NetApp maintenance contracts with a transparent, pay-as-you-go cloud cost model.
- Seamless Back-to-School Concurrency: Handled back-to-school registration rushes with 100% uptime and automatic scaling, eliminating seasonal server crashes.
Strategic Product Trade-offs
Active-Active Dual-Run vs. Accelerated Big-Bang Cutover
Early in migration planning, infrastructure stakeholders advocated for a single weekend "big-bang" cutover to compress the project schedule and avoid the engineering overhead of building a dual-run proxy.
From a product standpoint, we resisted the big-bang approach. With 200 million assets powering live classrooms and in-flight publication deadlines, a rollback failure during a hard cutover would have crippled editorial operations and disrupted millions of active students.
Instead, we invested the engineering effort into building an active-active dual-write middleware proxy paired with automated checksum verification. While this added initial development complexity and maintained temporary dual-storage costs for 90 days, it provided an indispensable safety net: zero downtime, continuous editorial productivity, and absolute proof of data parity before retiring legacy storage.
Lessons Learned
1. Byte-Level Verification Builds Unshakeable Trust:
When migrating millions of enterprise assets, anecdotal testing is insufficient. Encoding automated MD5 checksum validation into the pipeline gave editorial stakeholders complete confidence that their intellectual property remained intact.
2. Abstract Storage Interfaces Early:
Upstream authoring systems should never depend on physical storage paths. Wrapping storage behind unified RESTful API endpoints decoupled application logic from physical disk locations, ensuring future infrastructure upgrades require zero changes from content creators.
3. Dual-Run Overhead is Cheap Insurance:
The operational cost of running parallel environments for 90 days was negligible compared to the business and reputational damage of an unplanned production outage during fall semester registration.
4. Observability Turns Reactive Firefighting Proactive:
Implementing real-time APM telemetry and Grafana dashboards allowed our teams to detect latency degradation and edge cache misses before they impacted students, shifting our posture from incident triage to predictive tuning.
Conclusion
The Media Server cloud migration modernized Pearson's most critical digital asset repository, replacing brittle on-premises infrastructure with a resilient, globally distributed cloud architecture. By prioritizing business continuity, bit-level data integrity, and automated telemetry, the initiative eliminated operational bottlenecks and delivered a scalable foundation capable of supporting digital learning for millions of students worldwide.
Leading this project demonstrated that infrastructure modernization is fundamentally a product discipline: aligning engineering architecture with business goals, de-risking high-stakes operational transitions, and delivering seamless continuity for users.