Ask HN: A memory efficient and concurrent redis – are you interested? | Hacker News Reader

Ask HN: A memory efficient and concurrent redis – are you interested?

7 pointsqkhhly11 comments

We are working on a redis-compatiable key-value store. Our design reduced memory consumption by half for the same dataset compared with redis. We also implemented multi-threading and ACID compliance. In a 16-core machine, we achieved more than 10X throughput compared to redis. Our implementation is 100% compatible with redis clients (it can be used as a drop-in replacement of redis). We would like to gauge the interest in our product. Please leave a comment if you would like to try it out or if you have additional feature requests.

View on HN

I want to try as well. I'm curious, (if you don't mind sharing this information) what language is used to develop this new technologie ?

It's developed in C++. We have considered Go, but one of the major goal we want to achieve always non-blocking. Go can be blocked by GC, and redis maybe frequently blocked by fork/heavy commands.

Are there Jepson results?

https://jepsen.io/

Seems an interesting tool. The only question is how effective it is. But we should definitely try it out.

Over the past few years, Jepsen has developed a reputation within the database community for having an ability to verify claims about databases.

https://aphyr.com/tags/jepsen

it would be nice if there is possible to "move" values to somewhere else -- on other type of persistence in case of eviction, or simply write on a file in the simplest form -- just an idea

Does it include pub/sub?

We have implemented the core features. Pub/sub is on our todo list. Is pub/sub a critical feature to you? If we support pub/sub, what would be your motivation/obstacle to switch to our product?

I'm currently using redis in a pair of applications, one of which also uses pub/sub. I have many other applications which would benefit from the caching layer that redis provides, and since they currently have none there really isn't an obstacle to testing another solution.

I can try too.

I want to try

Reply on news.ycombinator.com