Rate Limits per Channel
What you'll learn: How to handle rate limits imposed by different notification providers (Apple, Google, SMS, email) to avoid throttling and service disruptions.
Every notification channel enforces rate limits—maximum requests per second, per user, or per app—to protect their infrastructure and prevent spam. Ignoring these limits leads to dropped messages, temporary bans, or degraded service.
Understanding Provider-Specific Limits
Apple Push Notification service (APNs):
- No hard per-second limit, but rapid bursts are discouraged
- Violations trigger connection throttling or temporary blocks
- Different priorities affect delivery speed
Google Firebase Cloud Messaging (FCM):
- ~240,000 messages per minute per project
- Per-device limits to prevent spam
- Topic-based messaging has separate quotas
SMS Providers (Twilio, AWS SNS):
- Typically 10-100 messages per second per account
- Vary by country and phone number type
- Cost-based throttling on shared pools
Email Services (SendGrid, SES):
- Start at ~1-14 messages per second
- Reputation-based scaling over time
- Per-campaign and per-recipient limits
Handling Rate Limits
Implement token bucket or leaky bucket algorithms per channel:
- Track available quota in real-time
- Queue excess messages for retry
- Distribute load across multiple API keys or sender pools
- Monitor provider-specific error codes (HTTP 429, service-specific rejection codes)
Use dynamic backoff strategies when limits are hit, and route urgent notifications through channels with available capacity.
Analogy
Think of each channel as a toll booth with different speed limits. APNs is a wide highway with flexible lanes, FCM is a multi-lane road with clear speed limits, SMS is a narrow bridge with strict capacity, and email is a freight lane that speeds up based on your driving history.
Key Takeaway:
Rate limits are channel-specific constraints that require per-provider quota tracking, intelligent queuing, and graceful degradation to maintain reliable notification delivery at scale.