Solution
Shopify's commerce platform processes record-breaking transaction volumes on Google Cloud during Black Friday weekend 2025, handling $14.6B in global sales and 489M requests per minute at peak. Compute Engine runs general-purpose commerce services on N2/N4 machines, C4 instances power Shop App catalog search, and Z-series machines provide high-performance local storage for real-time package tracking across 136M deliveries. Google Kubernetes Engine orchestrates auto-scaling microservices to absorb traffic spikes instantly. Hyperdisk Balanced and Extreme store databases supporting 14.8T queries and 1.75T writes during the weekend. AI workloads leverage accelerator-optimized VMs to power Shopify Magic recommendations and Sidekick assistant features.
Data flow
Customer purchase requests → N2/N4 Compute Engine processing → checkout service updates order database on Hyperdisk Extreme → payment processing → inventory deduction → Z-series machines track 136M packages globally in real-time via Shop App. Recommendation queries route to C4 machines for catalog search. AI inference for Shopify Magic and Sidekick routes to accelerator-optimized VMs. All responses aggregated through GKE service mesh.
Solution architecture
5 components · 4 layers - Compute
- Compute Engine Runs core commerce services on N2/N4 general-purpose machines, C4 compute-optimized machines for search, and Z-series storage-optimized machines for distributed database layers.
- Storage
- Persistent Disk Provides standard block storage for general application workloads and middleware components.
- Hyperdisk Balanced Delivers high-throughput, balanced performance storage for transactional databases supporting order processing and customer data.
- Orchestration
- Google Kubernetes Engine Orchestrates containerized microservices with instant horizontal auto-scaling to handle traffic spikes from 0 to 489M requests/minute during peak demand.
- Uncategorized
- Hyperdisk Extreme Enables ultra-high-performance databases for real-time inventory and package tracking handling 14.8T queries and 1.75T writes during peak shopping periods.
- Accelerator-optimized VMs powering Shopify Magic and Sidekick AI assistant
- Classic compute workloads on GKE call AI services through internal APIs; Hyperdisk provides high-throughput delivery
- Compute Engine N2 and N4 for general-purpose core commerce, C4 for Shop App catalogue search
- Custom machine types tuning CPU-to-memory ratio per workload
- Distributed global database layer on Compute Engine Z-series machines with high-performance local storage
- Microservices on GKE for instant elastic scaling during traffic spikes
- Persistent Disk for general storage; Hyperdisk Balanced and Extreme for high-performance databases
489M requests per minute at peak traffic
Processed $14.6B in global sales, a 27% increase YoY, during Black Friday weekend
Shopify has also built a distributed global-scale database layer on Compute Engine's Z-series machines, which provide high-performance local storage.
We chose Google because we needed a partner who could understand and support hyper-scale reliability engineering.
We used Google's accelerator-optimized VMs to take these workloads from PoC to production.
