Hello,

I love the builtin bandwidth-test in mikrotik, and now that i have a mikrotik CHR in the cloud to test to, I'm very happy with my ability to remote test speeds.

Customers love speedtest.net, of course.

Have any of you run into situations when the two measure dramatically different things?

I have a customer where i can login to their CPE and consistently test 20Mbps via mikrotik bandwidth-test to/from a cloud server on the internet. There is nothing special about the virtual server that is currently hosting the bandwidth-test..it's ip is not prioritized anywhere over any other Internet IPs.

When my tech tests from their location via speedtest.net (wired directly to the CPE, no other devices connected, wifi on the laptop off). They test 2-3Mbps. The customer sees the same results. They have tried testing to multiple different speedtest servers.

Any suggestions? I have tested a lot of connections in many different networks - and I have never run into an issue like this before.

Thanks much!

There could be some differences. For one, the clients behind the Mikrotik, do they go through any NATs or firewall rules, queues or anything else? That could slow down their speeds - depends on what the hardware is capable of.

Then you have general internet routing. What you want to do is get a traceroute from the CPE to your bandwidth testing device and then another to speedtest.net. Ensure that you are getting a traceroute to the server they are testing against. The route to the speedtest server could be longer or take a more congested route thus leading to poor results.

That brings me to the next possbility, they need to select a different test server. It is possible the one they are testing against is the bottleneck and not the Mikrotik.

The final possibility that I can think of off the top of my head is the customer's testing device, i.e. computer, is the problem.

Thanks for your reply Revelation:

Those are helpful suggestions, unfortunately I've thought of and tried many/most of them.

Regarding customer test device - The customer has done tests, and my tech has done tests on different devices. In particular, my tech's device is known good, and has successfully done many other speedtests at other locations that don't exhibit this problem.

Regarding speedtest.net server selection - There are a number of servers in the immediate geographic area. We have one we use as our standard, and it produces good results at other customer locations.

Regarding NAT / firewall / queue question - This is definitely the first thing i thought through / looked at. the CPE configuration is very simple in my network. The CPE is a PPPoE client, and all of the customer's traffic is NATted out the PPPoE interface.

Both my remote bandwidth-test tests and my tech's speedtest.net tests are exiting the CPE through the same PPPoE interface and are NATTed behind the same IP. This means for everything in my backbone and core setup both tests are passing through all the same queues and policies.

I only have one upstream connection to my core - and so both tests are routed through this. It is true that once the tests leave my core upstream interface that the routing is beyond my control...but the strange thing is that the cloud CHR router i am using as a target for bandwidth-tests is *much farther* geographically than all the speedtest.net servers.

I wonder how bandwidth-test and speedtest.net tests are functionally different in terms of the sessions and packets used to measure speed. Does anyone know much about the underlying bandwidth test methods for either tool? I'm trying to see if knowing that might shed some light on the subject.

Let's see the speed test command for the CHR test.

Speedtest is obviously using multiple tcp streams. Is your CHR test using udp or tcp? Udp throughput is usually greater as there is less overhead.

Sent from my SM-G920I using Tapatalk

Thanks Arcee -

Sometimes explaining things to others help you find problems yourself

I think this was the issue, I did not realize that bandwidth-test was defaulting to udp. I actually realized that before reading your post....once I started setting protocol=tcp and seeing very similar results to the speedtest.net results.

Thanks to that, I think I was able to narrow this down to really old firmware on a (non mikrotik) switch in the middle that had an issue with tcp in vlans. Going to have to keep testing, but I think i may have sorted this.

Hello,

We are having the same issue with on of our client's site. From Bandwidth Test we are continuously getting 25 Mbps but from client's PC or our PC we are not getting more than 10 Mbps. Checked all the suggestions already suggested by people on this forum. Please advise