1

Dear, i am using MongoDb atlas with M40 configuration.
When use M40 General, CPU is stability ~70% with abount 1300-1500 connections
but when switch to M40 SSD NVMe, CPU is up to 100% with the same traffic.

Anyone can support me about thís issue ???

thanks a lot.

2

May be, just may be.

Before disk I/O was a bottleneck. It is not anymore. Traffic is the same, but what about latency?

3

latencies were about 0.8 - 2 seconds

4

So, which choice is better, pls

5

one more thing i forgot, the Disk Util just about <= 10%, so i dont think it is disk I/O issue

6

smaller latency is better

7

10% in both cases? Or 10% in case of ssd? What was it with the old config?

8

10% in case of ssd, with general, old config, it was higher, about 30-40%

9

So

SSD => more CPU, less disk
Old HD => less CPU, more disk

We need to compare the latency of both. You gave a number but it is not clear for which. The specific number is also useless. The comparative values are important. The comparative values at different traffic load can be useful.

Are you maxed out in terms of traffic in any of the two cases?

10

I’ll will do more test and send you the more detail report, currently i did the test with CCU is 500, backend app with 20 threads, db is M40 5 nodes

As you said:
SSD => more CPU, less disk
Old HD => less CPU, more disk
So i understood a bit the issue.

11

A post was split to a new topic: M40 General vs NVMe SSD CPU count