Built into HotKV™
Built in, not bolted on.
HotKV speaks the Redis protocol and adds features of its own. Pick any of them below and watch an animated example of it working, with the real commands.
- protocol
- RESP2 + RESP3
- native code
- C++20
- Linux servers
- x86_64 + ARM64
- modules to install
- 0
See every feature work
Choose a feature. The demo types real HotKV commands with example replies, and shows what each one does inside the engine.
Key Tagging
Label keys with tags and query them: TAG.SET, TAG.GET, TAG.MEMBERS, TAG.INTER, TAG.COUNT, TAG.EXISTS, TAG.REM. No secondary index, no Lua script, no external service.
Keys get labels, and one command finds every key with a label, or with several labels at once.
Three ways to use a multi-core server
A server can execute commands on one main thread, split its data into a shard per thread, or let every thread work on one shared store. HotKV is built from scratch around the third design, so a single instance uses every core.
- I/O model
- Event loop + I/O threads for read/write
- Scaling
- Diminishing returns beyond I/O saturation
- Multi-core utilization
- Commands run on one thread; more cores need several instances or a cluster
- I/O model
- io_uring / epoll per thread
- Scaling
- Scales across cores; multi-key commands coordinate across threads
- Multi-core utilization
- Internal sharding per thread
- I/O model
- Multi-threaded, on every core
- Scaling
- Throughput scales with core count
- Multi-core utilization
- Single instance uses all cores
A schematic of how commands are executed, not a benchmark. Measured results are on the Performance page. Performance
Multi-threaded from the ground up
Each worker thread reads, parses and executes its own connections' commands, so I/O and execution both run in parallel and throughput grows with core count.
Predictable latency
A native engine with no garbage collector, so there are no stop-the-world pauses.
Modern C++20
HotKV is written in C++20 and compiled to native code.
Built for Linux servers
Packages for x86_64 and ARM64 Linux. Vector search uses the CPU's vector instructions where they are available.
Feature Comparison
How HotKV™ compares with Redis, Valkey, and Dragonfly, checked against each project's latest release.
Compared with Redis 8.10.2, Valkey 9.1.2, and Dragonfly 2.0.0, the latest releases in September 2026. Redis 8 also ships JSON, time series, and probabilistic data structures, and Redis offers LangCache and Agent Memory as separate services. Valkey 9.2, now a release candidate, adds server-side hot key tracking and opt-in fork-less snapshots. On HotKV Cloud, the AI features come with the Performance and Scale plans.
Works With Your Stack
HotKV™ speaks RESP2 and RESP3, so the Redis client library you already use connects to it. Point it at your HotKV hostname.
Compatible Tools
Standard Redis tools connect without changes (see the command reference for the commands HotKV supports)