- Newest
- Most votes
- Most comments
It's not defined because of a couple reasons:
- You can alter the MTU of any OS
- Many instances have burst credits (see here: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/burstable-credits-baseline-concepts.html)
The best way to see if an instance meets your needs is to configure the OS as you desire and then perform a sustained test to ensure you are not just seeing burst maximums.
So does that mean, the instance size has to be tested by trial and error to see if the PSS allowance is exceeded each time? Today I have the PPS during a time period when we started to see a service degradation. We are using two T3 medium instances behind an NLB and are seeing 125 (for T3 XL) and 86 (M instance size) of Packets per second using the basic metrics collected per instance in CloudWatch. What is the recommended way to scale up the instances such that there will be no service degradation? at the point of service degradation we see 37 K Packets and 25 K packets coming into the instance because the application must have started to retry!
Relevant content
- asked 5 months ago
- Accepted Answerasked 22 days ago
- Accepted Answerasked 4 years ago
- asked 5 months ago
- AWS OFFICIALUpdated 4 days ago
- AWS OFFICIALUpdated 2 years ago
- AWS OFFICIALUpdated 2 years ago
- AWS OFFICIALUpdated a year ago
If there is a metric then it must be a maximum allowed packets per second which is exceeded. Hence it is not a dynamic limit came from the bandwidth. It may be different for different instances, sure. Bandwidth is also vague defined for such burstable instances but it is well defined for other instances. I am not sure why pps limits aren't listed anywhere.
MTU alters packet size.