Tools / Redis / Redis vs Upstash
The in-memory data store — cache, queues, pub/sub.
Redis is an in-memory data store used for caching, streaming and messaging, session management, and search. Data lives in memory, which is why it's commonly placed in front of slower databases to serve reads and writes with sub-millisecond latency. Beyond the core key-value store, Redis now covers a set of adjacent workloads: vector search and semantic query for RAG and AI apps, agent memory (the site cites 86.5% accuracy at roughly $0.07 per session in its own Agent Memory Benchmark), LangCache for caching LLM responses to cut token spend, a feature store for real-time ML, leaderboards, and CDC-based data integration. It also handles classic queue and pub/sub messaging via streams. Deployment options span Redis Cloud (managed), Redis Software (on-prem), Redis Open Source (currently Redis 8.8), and managed offerings through AWS, Azure Managed Redis, and Google Cloud Marketplace. Tooling includes Redis Insight (a GUI for visualizing and debugging data), RIOT for data ingestion, and official client libraries for Python, Node, Java, Go, .NET, and others. Documented use cases include fraud detection, real-time decisioning, and session storage, with industry pages for financial services, retail, gaming, healthcare, and telco. The homepage names customers like Ulta (40% increase in revenue through a Redis-powered online experience), Superlinked (100+ vector search queries per second), Sony LIV (37% faster API responses with a 15% lower memory footprint), and Plivo (over a billion API requests per month).
Serverless Redis and Kafka with pay-per-request pricing.
Upstash is a serverless data platform built around Redis, with additional products for vector search, messaging, workflows, file storage, and cloud containers for AI agents. Databases are accessed over HTTP/REST as well as standard clients, which makes them usable from serverless and edge functions where persistent TCP connections aren't practical. Pricing is per-request with a free tier, and data is replicated across multiple regions worldwide.