🔴 To support my channel, I'd like to offer Mentorship/On-the-Job Support/Consulting (me@antonputra.com) 🍿 Benchmarks: th-cam.com/play/PLiMWaCMwGJXmcDLvMQeORJ-j_jayKaLVn.html&si=p-UOaVM_6_SFx52H 👋 AWS is expensive - Infra Support Fund: buymeacoffee.com/antonputra
What this proves is that dragonfly is its the easy way to achieve better performance with no effort vs a single redis instance and once you're getting close to the max usage you can put another instance up just like with redis for even more performance.
One huge performance gain when setting or retrieving data is to pipeline your commands or use server side lua scripts. It is far easier to do this when connecting to a single instance that is scaled vertically. To pipeline commands to a horizontally scaled setup is much more complicated.
The dragonfly claim still has some merit if your cached values don't have a uniform popularity. Scaling different masters won't help if one of the masters has the data everyone wants. (am i understanding correctly?)
Would love to see a comparison between some one of the redis/dragonfly and garnet. It's a new cache developed by microsoft that claims some crazy performance gains and is redis api compatible 🤔
Hi Anton, your channel is just a gem on TH-cam. Thank you for your amazing content. I have one question outside of the channel theme. What software do you use to record and add these amazing diagrams?
Thank you for your videos! They are very interesting and useful! It's very nice that your benchmark seems like real environment, because companies usually show, how fast their one instance compared with one redis instance when they run on one server with 64 cores, for example. But who would use redis this way 😅
Could you benchmark memcached vs dragonflydb with single multi-threaded instance. Also could you do it with each batch of around 20 keys per roundtrip instead. Because anyone who uses caching heavily know that batching requests improve throughout by a lot I remember Facebook once said they need to get more than thousand of keys from their cache just to render a single page.
С кешами у тебя пока неправильные тесты. dragonfly ориентирован на многоядер и 4 это не много. И redis можно складывать на полку, valkey его заменил. Но и они же могут лежать на полке, т.к. есть keydb. Ну и всех их можно убирать в стол, т.к. есть garnet
Please stop the auto generated audio translation. It's so annoying and bad quality. It is not possible to disable it globally. I skip almost all YT videos that start with it.
"This annoying "feature" was implemented a long time ago, the bad part is that there is no way in the settings to disable it, except from changing your UI language to English."
🔴 To support my channel, I'd like to offer Mentorship/On-the-Job Support/Consulting (me@antonputra.com)
🍿 Benchmarks: th-cam.com/play/PLiMWaCMwGJXmcDLvMQeORJ-j_jayKaLVn.html&si=p-UOaVM_6_SFx52H
👋 AWS is expensive - Infra Support Fund: buymeacoffee.com/antonputra
Redis vs KeyDB vs Valkey
Also I would like 🎉
vs Kvrocks
Vs skytable
And maybe vs tarantool 😅
will do!
¡Thanks!
thank you for the support!!
Love it cause from the thumbnail i have an idea of which of them perform better, thanks you for the valuable content Anton
Great comparison once again, thank you! Can we have Nats vs Kafka please?
yes i'll do nats soon
@@AntonPutrathanks
What this proves is that dragonfly is its the easy way to achieve better performance with no effort vs a single redis instance and once you're getting close to the max usage you can put another instance up just like with redis for even more performance.
true
Well for real performances with Redis you need a cluster and set it up
One huge performance gain when setting or retrieving data is to pipeline your commands or use server side lua scripts. It is far easier to do this when connecting to a single instance that is scaled vertically. To pipeline commands to a horizontally scaled setup is much more complicated.
interesting
🚀 Redis vs Garnet (no sure why no one asking for this)
ok added
I think Garnet vs Dragonfly will be a more fair comparison, but yes, I also want to see a Garnet benchmark.
Thank you, for such a great comparison!
my pleasure!
Hey can you do:
ScyllaDB vs MongoDB vs SurrealDB
SureealDB isn't quite the same as the others, but nonetheless I'm curious as well
The autodubbed audio tracks are just bad. Why is youtube automatically choosing the wrong audiotrack?
it takes few hours to process i guess
Skytable is also pretty fast, if I remember correctly the fastest, hope you try that as well
ok interesting
Garnet comparison please!
ok added
The dragonfly claim still has some merit if your cached values don't have a uniform popularity. Scaling different masters won't help if one of the masters has the data everyone wants. (am i understanding correctly?)
redis keys usually spread evenly across all the masters, you can see it from the cpu usage graph, even distribution
hi, are you planning to compare Redis vs KeyDB?
yes Redis vs KeyDB vs Valkey in a week or so
If you gave Redis also more cores, you could enable the io-threads functionality in Redis for a better son multi-threading.
Timeseries Database comparison as they are often used in high throuput ingestion scenarios.
InfluxDB vs TimescaleDB vs QuestDB 📈
I would love to see how intensive logging and tracing affects performance
noted
Would love to see a comparison between some one of the redis/dragonfly and garnet. It's a new cache developed by microsoft that claims some crazy performance gains and is redis api compatible 🤔
ok added garnet to my list
Would you scale kv store based on cpu usage? Doesn't the cpu monitoring issue make dragonfly undeployable?
i use cpu usage for capacity planning, so it's a very important metric for me. there are some workarounds for dragonfly
Your videos are gold! Can you please compare SQLite vs rqlite vs libsql (turso)?
They are the most popular embedded databases.
Please Garnet vs some other like redis etc.
got it
Hi Anton, your channel is just a gem on TH-cam. Thank you for your amazing content. I have one question outside of the channel theme. What software do you use to record and add these amazing diagrams?
Thank you for your videos!
They are very interesting and useful!
It's very nice that your benchmark seems like real environment, because companies usually show, how fast their one instance compared with one redis instance when they run on one server with 64 cores, for example.
But who would use redis this way 😅
If you need just cache, just use memcached
yes!
Did you optimize the kernel for redis ?
Could you benchmark memcached vs dragonflydb with single multi-threaded instance.
Also could you do it with each batch of around 20 keys per roundtrip instead. Because anyone who uses caching heavily know that batching requests improve throughout by a lot
I remember Facebook once said they need to get more than thousand of keys from their cache just to render a single page.
thanks for the feedback, i may do batch in the future
1 instance vs 4?)
4 cpu vs 1 + 1 + 1 + 1
Please do Litespeed vs Caddy!
Redis vs Garnet
garnet will be a great choice
noted!
Kafka vs Redpanda please
С кешами у тебя пока неправильные тесты. dragonfly ориентирован на многоядер и 4 это не много. И redis можно складывать на полку, valkey его заменил. Но и они же могут лежать на полке, т.к. есть keydb. Ну и всех их можно убирать в стол, т.к. есть garnet
Apache Kafka vs Apache Pulsar
RabbitMQ vs LavinMQ
Redis vs Kvrocks
ok
People been asking to compare elixir for so long now, please ❤
yes, the next video is to find the fastest go framework and right after, elixir vs. go
@@AntonPutra you are the best
@@AntonPutraspicey!
Redis vs DiceDB
Is your face AI-generated?
no
Maybe 2x m7a.large valkey performs better than 1x m7a.xlarge? scaling both vertically and horizontally sounds like a sane approach here
i'll test few configurations for valkey
WHAT?
Please stop the auto generated audio translation. It's so annoying and bad quality. It is not possible to disable it globally. I skip almost all YT videos that start with it.
ok let me see if i can disable it
"This annoying "feature" was implemented a long time ago, the bad part is that there is no way in the settings to disable it, except from changing your UI language to English."