7

I was going through benchmark documentation page of Redis and Etcd. From the benchmark data it seems Etcd is as efficient as Redis.

https://redis.io/topics/benchmarks

https://github.com/etcd-io/etcd/blob/master/Documentation/op-guide/performance.md#benchmarks

Example - Setting a 100k keys by 100 parallel clients with payload - 256 byte

- on redis : 70K QPS 
- on Etcd : 50k QPS

But I am not able to understand how come Etcd is as efficient as Redis or is there anything wrong in my understanding?

Reason for Etcd should have been much slower than Redis:

  • Etcd might be making use of SSD but still Redis is in-memory DB, so should have been much high performant.
  • Etcd provide strong consistency using consensus(Raft) and should have been slower than Redis. As Redis doesn't guarantee consistency.
Rahul
  • 326
  • 2
  • 10
  • you have to take into account the benchmark setup as well. – Tuan Anh Tran Aug 26 '20 at 09:18
  • Agree. That might add up but still it shouldn't be of much difference. System used for etcd is - 8 vCPUs + 16GB Memory + 50GB SSD – Rahul Aug 26 '20 at 09:55
  • redis version in the post is 2.4. Current version is 6. Redis 2.4 can do `SET: 198412.69 requests per second` which is a lot higher than your 70k. btw, it's better to do a benchmark on your own workload to see which one is best for you – Tuan Anh Tran Aug 26 '20 at 10:03
  • It is linear growth (4 times) with high end server on unix socket and not TCP loopback. I am not sure if there is huge performance improvement in redis 6 over 2.4 (for write). – Rahul Aug 26 '20 at 10:34
  • 1
    btw, redis raft is coming too :D https://github.com/RedisLabs/redisraft – Tuan Anh Tran Aug 27 '20 at 02:12

0 Answers0