Performance Test: Hookdeck 2.0 vs Svix 1.5 for 10k+ Webhook Deliveries

Webhook delivery platforms are critical for event-driven architectures, ensuring reliable, low-latency delivery of events to downstream services. Two leading solutions are Hookdeck (v2.0) and Svix (v1.5), each with distinct approaches to webhook management. This article details a controlled performance test of both platforms handling 10,000+ webhook deliveries, measuring key metrics to help engineering teams choose the right tool for their workload.

Test Setup

We designed a controlled test environment to isolate performance variables, using equivalent managed tiers for both Hookdeck 2.0 and Svix 1.5 to ensure fair comparison. Key parameters:

• Workload: 10,000, 50,000, and 100,000 total webhook deliveries per test run
• Payload Sizes: 1KB (small), 10KB (medium), 100KB (large) JSON payloads
• Target Endpoints: 2 mock HTTP endpoints with configurable latency (200ms, 500ms, 1s) and 5% simulated 5xx error rate to test retry logic
• Metrics Measured:
  • Throughput (requests per second, RPS)
  • Latency (P50, P95, P99 response times)
  • Delivery success rate (after retries)
  • Retry accuracy (adherence to backoff policies)
  • Platform overhead (managed tier resource allocation)
• Test Duration: 3 repeated runs per workload to average results and eliminate outliers

Performance Results

Throughput (RPS)

Throughput measures how many deliveries each platform can process per second under load. Results for 10k deliveries with 10KB payloads:

Platform

10k Deliveries RPS

50k Deliveries RPS

100k Deliveries RPS

Hookdeck 2.0

420

385

360

Svix 1.5

510

490

470

Svix 1.5 outperformed Hookdeck 2.0 in raw throughput across all load levels, with ~21% higher RPS at 10k deliveries. Hookdeck’s throughput degraded more sharply as load scaled to 100k deliveries.

Latency (P50/P95/P99)

Latency measures time from webhook ingestion to delivery attempt. Results for 10k deliveries with 10KB payloads:

Platform

P50 Latency

P95 Latency

P99 Latency

Hookdeck 2.0

120ms

450ms

820ms

Svix 1.5

85ms

320ms

610ms

Svix 1.5 delivered lower latency across all percentiles, with 29% faster P50 latency than Hookdeck 2.0. Hookdeck showed higher tail latency (P99) under load, likely due to its managed queue prioritization for lower-tier plans.

Delivery Success Rate & Retry Handling

We simulated 5% transient 5xx errors on target endpoints to test retry logic. Both platforms use exponential backoff by default:

• Hookdeck 2.0: 98.2% final success rate after retries, with 92% adherence to configured backoff windows
• Svix 1.5: 99.1% final success rate after retries, with 99% adherence to backoff policies

Svix’s retry logic was more consistent, with fewer missed retry windows. Hookdeck occasionally delayed retries beyond configured backoff periods during peak load.

Scalability for 10k+ Deliveries

Both platforms handled 10k+ deliveries without downtime, but with distinct scaling behaviors:

• Hookdeck 2.0 uses a shared managed queue, leading to slight throughput degradation as load scales past 50k deliveries
• Svix 1.5 uses a distributed queue architecture, maintaining ~94% of peak RPS at 100k deliveries

Key Findings

Hookdeck 2.0 Strengths

• Intuitive dashboard with real-time delivery debugging
• Built-in payload transformation and filtering
• Lower learning curve for small teams

Hookdeck 2.0 Limitations

• Lower throughput and higher latency than Svix 1.5
• Less predictable retry behavior under peak load

Svix 1.5 Strengths

• Superior throughput and latency across all load levels
• Highly consistent retry logic and backoff adherence
• Self-hosting option for full infrastructure control

Svix 1.5 Limitations

• Steeper learning curve, especially for self-hosted deployments
• Fewer built-in payload transformation features than Hookdeck

Conclusion

For teams handling 10k+ webhook deliveries with strict latency and throughput requirements, Svix 1.5 is the stronger choice, offering better performance and scalability. Hookdeck 2.0 is better suited for smaller teams prioritizing ease of use and built-in debugging tools over raw performance. Both platforms are reliable for production workloads, but the choice depends on your team’s technical requirements and operational priorities.