Case Study

Poizon (also known as Dewu) is China's leading sneaker and streetwear marketplace, renowned for its product authentication services. The platform's observability infrastructure—collecting logs, metrics, and traces from millions of transactions—is critical to maintaining trust and service quality. Managing this at scale presented significant challenges:

• Extreme Traffic Peaks: Flash sales and product drops generate observability data at rates exceeding 40 GiB/s. Traditional Kafka required massive over-provisioning to handle these unpredictable spikes, leading to wasted resources during normal operations.
• Storage Cost Explosion: Observability data has long retention requirements (weeks to months) for troubleshooting and compliance. Storing this on expensive local disks attached to Kafka brokers was financially unsustainable.
• Scaling Rigidity: The legacy architecture couldn't adapt to traffic patterns. Scaling up was slow due to partition rebalancing, and scaling down was avoided altogether to prevent data loss—resulting in perpetual over-provisioning.

Poizon chose AutoMQ to build an observability backbone that could breathe with their business.

• Elastic Compute Scaling: AutoMQ's stateless broker design allows the compute layer to scale independently. When traffic surges during a flash sale, new brokers spin up in seconds. When traffic subsides, they spin down automatically. This elasticity, previously impossible with stateful Kafka, is the key to cost efficiency.
• Tiered Storage for Cold Data: AutoMQ natively supports tiering data to object storage (OSS). Recent, "hot" data is served from fast storage, while older, "cold" data is automatically moved to OSS at a fraction of the cost—up to 85% savings for long-term retention.
• Observability Optimized: The platform is purpose-built for high-throughput log and metrics ingestion. Its architecture handles the bursty, write-heavy nature of observability workloads gracefully.

An Observability Platform That Scales with the Business

The migration to AutoMQ has delivered a transformative outcome for Poizon's observability infrastructure.

• Peak Performance: The new architecture handles 40 GiB/s observability workloads stably during flash sales, with headroom to spare. The team no longer worries about the platform buckling under pressure.
• Dramatic Cost Savings: Elastic scaling alone reduced compute costs by approximately 50%. Tiered storage for long-term retention slashed storage costs by up to 85%. Combined, the total cost of ownership for the observability platform has been fundamentally reduced.
• Operational Simplicity: Scaling is now automated and event-driven. The team defines scaling policies, and AutoMQ handles the rest. The operational burden of managing stateful Kafka during high-stakes traffic events is gone.