Serverless Redis and Kafka platform with pay-per-use pricing, global deployment, and zero ops — the go-to caching and message queue solution for serverless applications.
Because access goes through HTTP rather than the native Redis protocol, Upstash latency is 5-20ms higher than self-hosted Redis. For scenarios requiring sub-millisecond responses (games, real-time bidding), this difference is fatal. But for most web apps (caching, sessions, rate limiting), it's fully acceptable.
10K commands/day sounds like a lot, but a single page load can trigger 3-5 Redis queries (session + cache + rate limit). With 500 DAU, that's 2,500-5,000 commands per day — adding one background task could push you over. Once exceeded, you get throttled rather than auto-upgraded to the paid tier.
QStash guarantees at-least-once delivery — but network retries can result in the same message being delivered twice. Your consumer must be idempotent (processing the same message twice produces the same result). This is not a QStash bug — it's common to all message queues — but is frequently overlooked by newcomers.
$0.2/100K commands looks cheap, but if your OPC product is data-intensive (e.g., querying Redis 50 times per request), costs can accumulate quickly. At 1M requests/day, Upstash can be far more expensive than self-hosted Redis. Do a cost estimate before deciding.
| Rate Limiting in Next.js with Upstash | Blog | - |
