The Call That Changed Everything
I was running the Data Center practice for ePlus in the Western United States, serving Silicon Valley giants like Google, Oracle, and NetApp. We had carved out a unique position in high-tech, specializing in data center networking, systems, and storage. My engineering practice covered the entire West Coast.
Then I got the call from my friend Ed, who was running infrastructure for Apple iTunes at the time:
“Colin, we need the largest content delivery network on Earth built in six months. Will you come and help us?”
I said yes.
The Challenge: 6 Months to Build the Impossible
Apple needed to build the largest content delivery network on Earth—and they needed it operational in six months. This wasn’t incremental growth; this was building for a billion users from day one.
I brought together a coalition of the best:
- ePlus (my company) for integration and architecture
- Cisco Systems for switching and routing, including CRS-1 carrier routers
- Juniper Networks for firewalling and security
- Multiple technical partners for specialized components
Revolutionary Architecture: First of Its Kind
I architected and deployed the world’s first production multi-chassis link aggregation (MLAG) architecture at this scale. This was bleeding-edge technology that became the standard for hyperscale data centers.
My Responsibilities as Network Architect:
- Data Center Switching: Designed the entire Layer 2 switching fabric
- Data Center Routing: Built the internal routing architecture
- Edge Routing: Deployed Cisco CRS-1 carrier routing systems—the largest routers on Earth at the time
- Security Architecture: Implemented comprehensive Juniper firewalling across all data centers
- Software Development Teams: Led the teams building automation and orchestration
The Build: Maiden, North Carolina
We deployed what was then the largest data center on Earth:
- Location: Massive facility in Maiden, North Carolina
- Investment: $113 million in network gear alone
- Timeline: 6 months from design to operational
- Scale: Capacity to serve every iTunes/iCloud user globally
The Maiden facility was a marvel of engineering—strategically located for cooling efficiency and connectivity to both US and European markets.
Innovation Under Extreme Pressure
Construction Challenges
With construction crews racing to build the physical infrastructure, we built software so that the construction workers only had to focus on screwing things in. Everything else was built from code. This revolutionary approach meant:
- Zero-touch provisioning: Equipment configured itself when powered on
- Automated validation: Software verified every connection and configuration
- Self-documenting infrastructure: The code became the documentation
- Construction crews could focus purely on physical installation
Technical Breakthroughs
The multi-chassis link aggregation architecture I designed allowed:
- Horizontal scaling without service interruption
- Redundancy at every layer without complexity
- Performance that exceeded Apple’s aggressive requirements
- A foundation that could grow 100x without redesign
Lessons from the Edge of Scale
This project taught me fundamental truths about building at hyperscale:
Technical Lessons
- Service Router Architecture: Took lessons from service routers and applied them to data center design
- Chipset Design Patterns: Applied high-performance compute cluster patterns from chipset designers
- Multi-Chassis Innovation: Proved MLAG could work at scales others thought impossible
- Automation at Scale: Built software development teams that automated what would have required hundreds of operators
Operational Lessons
Working with construction crews taught me:
- How to deploy infrastructure in parallel with construction
- The importance of modular, repeatable designs
- How to train teams on technology that doesn’t exist yet
- The value of clear documentation when you’re inventing as you build
The Ripple Effect
This work became pivotal for the entire industry:
Industry Impact
- Set the Standard: Our MLAG architecture became the template for hyperscale data centers
- Enabled HPC Evolution: Patterns we developed were later adopted for high-performance compute clusters
- Influenced Giants: Oracle, chipset designers, and others adopted our architectural patterns
- Changed the Game: Proved that content delivery at this scale was not just possible, but reliable
Personal Impact
- Learned to design for scales that seemed impossible
- Developed techniques for managing $100M+ technical deployments
- Built relationships with the world’s best infrastructure teams
- Gained experience that would later prove invaluable at Oracle and beyond
From iTunes to AI
The principles I learned building Earth’s largest content delivery network now guide my AI infrastructure work:
- Start with the Impossible: If it seems achievable with current tech, you’re not thinking big enough
- Architecture Beats Hardware: The right design can make commodity hardware outperform exotic systems
- Automate from Day One: Manual processes that work at small scale will kill you at large scale
- Build Coalitions: No single vendor has all the answers—orchestrate the best from each
The Bottom Line
In six months, we built what others said would take two years. With $113 million in network gear and a small team of the best engineers on the planet, we created infrastructure that would serve billions of users for over a decade.
When Ed called asking for help building the largest content delivery network on Earth, he wasn’t exaggerating. We didn’t just meet the challenge—we redefined what was possible in data center networking.
Note: All information shared is based on publicly available information and non-confidential aspects of the project.