question

wijjo-0110 avatar image
0 Votes"
wijjo-0110 asked AndreasBaumgarten commented

A question about testing on Network Bandwidth/Throughput between two B1ms & two B2ms VMs

hello Support, i'm doing testing on Network Bandwidth/Throughput between two B1ms & two B2ms VMs in various locations arounf the world, using iperf3 tool.
in the testing result i got that the Bandwidth with B1ms VMs is better than B2ms VMs !!!
the B2ms VMs Bandwidth supposed to be much better, since B2ms is higher in performance CPUs&RAMs!!

Will you please help/clarify our confusion in this case?

Thank you in advanced.

azure-virtual-network
· 1
5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

Hi @wijjo-0110 ,

does the answers given so far work for you?

If you found the answer helpful, it would be great if you please mark it "Accept as answer". This will help others to find answers in Q&A


Regards
Andreas Baumgarten

0 Votes 0 ·
AndreasBaumgarten avatar image
0 Votes"
AndreasBaumgarten answered wijjo-0110 edited

Hi @wijjo-0110 ,

maybe the answer here from GitaraniSharmaMSFT-4262 is helpful:

https://docs.microsoft.com/en-us/answers/questions/52560/what-is-the-expected-network-bandwidth-for-vm-stan.html

Quote:

As per the Azure Compute backend team, the B-series is variable, so they have not published any networking numbers that may not always be accurate. The design of the B series does not provide the user with a consistent level of network performance. In addition, the B series is designed for more CPU intensive workloads as that is what we are really scaling out depending on credits. If network bandwidth is an issue then the B series is likely not the best choice.
You can find more information in the following thread : https://github.com/MicrosoftDocs/azure-docs/issues/8765

Even if a B2ms VM has a higher performance than a B1ms your measures depending on some variable facts. For instance the available burst credits at the time or your tests.


(If the reply was helpful please don't forget to upvote and/or accept as answer, thank you)

Regards
Andreas Baumgarten

· 1
5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

Dear AndreasBaumgarten ,

  1. thank you for your reply and the link, i do appreciate it.

  2. i have a scenario with 2 B1ms VMs and 2 B2ms VMs, all of them have the same configuration and have been tested at the "same time" using iperf3.
    and the results show a better Bandwidth for the B1m. as shown in the following table:

82216-b1ms-b2ms.jpg

it is not clear for me, how come the B1ms has a better BW than B2ms ? it is illogical for me!!
and what are the variable facts exactly (on Azure) that effect on the BW ?


thank you in advanced.


0 Votes 0 ·
b1ms-b2ms.jpg (29.6 KiB)
AndreasBaumgarten avatar image
0 Votes"
AndreasBaumgarten answered

Here are 2 links with the details how the B-Series VMs are working:
https://singhkays.medium.com/understanding-the-azure-b-series-and-cpu-credits-cd6ad1c46094
https://www.danielstechblog.io/monitoring-azure-b-series-vm-banked-credits/

The variable facts I am referring to is the CPU performance. Depending on the burst mode/burst credits of an individual VM the CPU performance is variable.
https://docs.microsoft.com/de-de/azure/virtual-machines/sizes-b-series-burstable

The CPU performance is influencing everything in the VM ... OS, running applications, drivers, disk performance, network performance, ...
This might be the reason why you get the results during your tests.

At least that's my idea. Maybe someone from Microsoft has some more insights.


(If the reply was helpful please don't forget to upvote and/or accept as answer, thank you)

Regards
Andreas Baumgarten

5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.