Amazon ElastiCache now supports Valkey 8.0, aws.amazon.com/blogs/databa.... This is the first release with software that was developed by the new Valkey project, which includes a wide variety of improvements, from performance to memory efficiency. I have two specific anecdotes to share about it. 🧵 - ThreadSky

reconditerose.bsky.social • 102 days ago

Amazon ElastiCache now supports Valkey 8.0, https://aws.amazon.com/blogs/database/amazon-elasticache-version-8-0-for-valkey-brings-faster-scaling-and-improved-memory-efficiency. This is the first release with software that was developed by the new Valkey project, which includes a wide variety of improvements, from performance to memory efficiency. I have two specific anecdotes to share about it. 🧵

Comments

george.black•98 days ago

Love reading threads like this, even if I don't understand half of it!

reconditerose.bsky.social•98 days ago

If you're interested in a less abridged version, you might find https://valkey.io/blog/unlock-one-million-rps/ and https://valkey.io/blog/valkey-memory-efficiency-8-0/ helpful.

george.black•98 days ago

Will check this out, thank you!

Funny enough – I just picked up Valkey & ElastiCache for the first time yesterday (for a side project). Never used Redis before either.

The simplicity of it (from the user's perspective) made it pretty fun!

reconditerose.bsky.social•102 days ago

One of the key features of Valkey 8.0 was dramatically improved multi-threaded performance, but ElastiCache already had that, AWS contributed it after all. What's exciting here is the IO threads in Valkey 8.0 could be re-allocated instantly! You can go from 150k to 1.2m RPS with a config change!

reconditerose.bsky.social•102 days ago

This was a huge win for our serverless offering, which nows supports 5x faster scaling, because we can now dynamically re-allocate threads to better utilize the underlying hardware so faster, that customers don't see latency spikes. A great win for open source and AWS customers!

reconditerose.bsky.social•102 days ago

The second curiosity is the improvements in memory efficiency. The AWS blog highlights 32 bytes of improvement, but that is higher than the OSS Valkey blog which says 23. The other 9 bytes actually come from an earlier improvement in Redis OSS, which AWS couldn't merge into our scaling system.

reconditerose.bsky.social•102 days ago

The reason was that we needed a way to iterate over all the keys in a "slot "(a logical shard of a Valkey cluster) and keep track of whether we had already processed a key. The system in Redis OSS 7.0 had no way to do that though, it only supported listing all the keys in the slot.

reconditerose.bsky.social•102 days ago

So we innovated! There was an OSS proposal that both improved memory efficiency and gave us the iteration properties we were looking for, so we implemented it for the community and then enabled it for our service. This is why we were able to pass on so much cost savings with Valkey, innovation!

Comments

Posting Rules

Reply