Media System Modernization

Overview

Sweetwater's product catalog features tens of thousands of items and millions of individual product images — including unique photo sets for every serialized guitar and high-value instrument.

Historically, all of those images were hosted on-premise in a cluster of load-balanced virtual machines. Images were imported via a legacy batch-processing server that ran hundreds of scheduled cron jobs. The result was growing storage constraints, rising maintenance cost, and limited scalability during traffic spikes.

When I took over leadership of the media services development pod, we set out to replace the legacy on-prem image pipeline with a modern, cloud-native architecture built for performance and infinite scale.

The Challenge

•Hundreds of nightly cron jobs running on a single server to sync images between systems
•On-prem storage nearing capacity due to multiple resized copies of every image
•Latency and reliability issues under peak traffic loads
•No autoscaling or intelligent caching
•Manual interventions required when syncs failed

Sweetwater's catalog contained about 60,000 active SKUs, each with 10–12 images — plus 5,000–8,000 serialized instruments photographed individually. This represented millions of image files and a massive amount of redundant data.

The Approach

Under my management, the team rebuilt the media system as a cloud-native Laravel service running on Google Kubernetes Engine (GKE).

Architecture:

✓Laravel API serving images directly from Google Cloud Storage (GCS)
✓Fastly CDN front-end with Image Optimizer for real-time resizing and compression
✓Horizontal Pod Autoscaler (HPA) configured for CPU and memory thresholds to handle variable traffic
✓Datadog used for monitoring, alerting, and performance visualization

Process Improvements:

✓Migrated all product and article images to GCS
✓Rewrote internal services to generate and serve URLs dynamically through the new media API
✓Partnered with Platform Engineering to tune scaling rules for seasonal demand surges

Results

500KB → 20KB

Average Image Payload (WebP)

Infinite

Scalability with Cloud Storage

Millions

Images Migrated to GCS

✓Performance: Cut average image payload from 500–800 KB to 20–50 KB using WebP compression
✓Scalability: Eliminated on-prem storage limits; new system scales elastically with traffic
✓Efficiency: No more pre-generated image sets — a single master file now serves every device size dynamically
✓Reliability: Reduced downtime and sync failures through autoscaling and centralized observability
✓Speed: Improved page-load performance site-wide by reducing image latency and optimizing delivery

Technology Stack

LaravelGKEGoogle Cloud StorageFastly CDN (Image Optimizer)DatadogKubernetes HPARedisWebP Compression

This project transformed image delivery from a fragile, on-prem bottleneck into a modern, self-scaling system that supports Sweetwater's massive catalog and future growth.

David Stillson

Engineering Lead, Sweetwater

Key Takeaways

✓Cloud-native architecture eliminates infrastructure bottlenecks and enables true scalability
✓Modern image formats (WebP) combined with CDN optimization can reduce payloads by 90%+
✓Dynamic image transformation eliminates the need for pre-generated image sets
✓Kubernetes HPA ensures cost efficiency while automatically handling traffic spikes
✓Comprehensive monitoring (Datadog) provides visibility and enables proactive issue resolution
✓Migrating from batch processing to API-driven architecture improves reliability and flexibility

Technologies Used

Overview

The Challenge

The Approach

Architecture:

Process Improvements:

Results

Technology Stack

Key Takeaways

More Case Studies

Marketplace Integrations

WordPress Rebuild & Optimization

B2B Commerce Transformation

Ready to Transform Your Business?