We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
Lobsters·Jun 11·5 min read
How We Moved Discord Voice to the Edge
TL;DR
Discord faced latency issues for users far from its voice servers, which were limited to about 30 global locations. To address this, they migrated voice and video traffic to Cloudflare's edge network, which has over 300 locations.
✦ Why It Matters
Engineers can leverage edge computing to reduce latency and improve user experience in distributed applications.
Key Takeaways
- Consider edge networks for latency-sensitive applications.
- Monitor quality metrics to assess performance improvements.
- Optimize software for distributed environments.
- Evaluate user experience across different geographic regions.
Full Summary
Discord's voice communication relied on a limited number of data centers, causing latency for users in remote areas. To enhance user experience, Discord migrated its voice and video traffic to Cloudflare's edge network, which features smaller points of presence (PoPs) in over 300 cities worldwide.
This migration involved building and optimizing their software to run efficiently across these numerous locations. The team monitored quality metrics, discovering significant improvements in user experience, particularly in Frankfurt, where average ping times decreased by 34% and packet loss dropped by 42%.
This shift not only improved performance but also allowed Discord to better serve users in regions previously underserved by major cloud providers. The findings underscore the importance of network architecture in real-time communication applications.