Launching today
KiviDB
The in-memory vector database for AI and real-time apps
5 followers
The in-memory vector database for AI and real-time apps
5 followers
A multithreaded, Redis- and Valkey-compatible in-memory database written in Rust, with full-text search, vector search and JSON built in. An alternative to Redis, Valkey and Dragonfly.
Free Options
Launch Team
AWS Startup AdvisorPersonalized AWS guidance that meets you where you’re already building
Promoted
Hi Product Hunt 👋
KiviDB is a Redis-compatible database that runs commands on every core instead of one. Your Redis clients and code work as they are; you change the connection string.
We benchmarked 1.2.0 on AWS Graviton3 bare metal (c7gn.metal, 32 cores), against Redis, Dragonfly and Garnet on the same machine:
• 21.2M GETs a second, 9× Redis 8.10, at a p99 of 0.36 ms (Redis: 2.29 ms)
• 17.8M SETs a second, 9.2× Redis
• 2.3× Dragonfly on pipelined reads
• 143 bytes a key, against Redis’s 178
We don’t win everything. Garnet is faster on pipelined writes and Dragonfly is slightly ahead without pipelining.
Both are in the report, and the scripts and raw results are on GitHub so you can rerun them yourself.
KiviDB Cloud runs it for you on AWS, Google Cloud or Azure. A replicated 5 GB database is $96 a month, 33–63% less than ElastiCache, Memorystore, Dragonfly Cloud or Redis Cloud. It’s billed by the hour, and you can try a free 100 MB database for a week without a card.
📄 Full report: kividb.io/benchmarks