MS220-8P Throughput Wrong

[svetlinski]

Hi,

I am not sure if anyone else has had the same issue?

I have a switch connected on port 8 to the router that provides Internet (TalkTalk). The same switch is connected on port 1 to a MR33.

When I use the dashboard and check the throughput, the AP returns 30Mbps (average) and the switch returns 9Mbps (average). 

I logged this with Meraki, but I feel I was fobbed off. The engineer said the below and closed the case. 

"This test is designed to test the speed of communication with Dashboard. It is not meant to accurately test the speed an Internet/WAN link is capable of."

I provided more information and re-opened the case, but the same engineer replied:

"I wouldn't put too much trust on this tool, it is basically used to test connectivity to the cloud, it won't give you the actual bandwidth as you may think."

I understand what he's saying, but how is it possible that the switch, directly connected to the router will get lower throughput than the AP that has to go through the same switch?

• The router's ports are only 100Mbps, but I don't think that would be of much difference here, as the speed of the Internet circuit is around 36Mbps? 
• The endpoint connected to port 6 gets 36Mbps when using Speedtest.net 

I think the support engineer wants to keep this buried, but in fact, I believe I should ask for another engineer's opinion?

Do you have any ideas/advice? What would you do?

All comments are appreciated!

Thanks in advance, guys!

Kind Regards,

Emanuil Svetlinski

[BlakeRichardson]

@svetlinski  The reporting will show ALL data, so port 1 will show traffic between the wrieless devices and devices on your LAN as well as to the internet. 

Port 8 since it has your router connected and I am assuming nothing else is plugged into the router will only show internet traffic. 

Its completely normal for ports that have clients devices on them to have much higher bandwitdth use than a port which has a router connected on it. 

[svetlinski]

@BlakeRichardson- do you mean the Summary Report on the dashboard?

I have already checked there and I seem to have a lot of data when I select the AP under the Networks dropdown, but if I change to the switch, I have no data at all...

I understand that you're saying about bandwidth, but should this affect the throughput of the switch to the Meraki cloud?

[PhilipDAth]

APs often have gruntier CPUs than switches, as they have a lot more processing to do.  The mathematics for AP processing are intense.

Switches have silicon to do all the heavy lifting, so the CPU doesn't do that much.  The CPU is the thing that does the speedtest.  The CPU in the MS220-8P (pretty much the lowest end switch Meraki has available) wont be terribly powerfull.

I don't know the answer for certain, but this may just be a reflection of the extra CPU power of the MR33.

[svetlinski]

I have already tested by unplugging the AP and testing. I also tested by disconnecting the devices from the switch and plugging them directly into the router. I had the same result. It makes no difference whether the AP is plugged into the switch or not, the switch always shows around 9Mbps throughput...

I just find it strange and I can't believe Meraki won't provide a proper explanation or a reason for this. Even if it is down to the slower CPU, surely throughput is throughput... I mean, when there were no other devices connected to the switch, technically, there should be not that much load on the CPU, right? I understand it still has to process the data from the router, but using only 1 port and still getting the same throughput? If it was down to CPU, then we should at least see the throughput increase significantly? Or am I wrong?

Sorry, just thinking out loud 🙂 

[PhilipDAth]

> If it was down to CPU, then we should at least see the throughput increase significantly? Or am I wrong?

No I don't think so.  I don't know, but I'll take a punt that it has a small ARM CPU, as they are low cost and you can get them SOC style - which is perfect for a small switch.

The traffic for the speedtest is sent over an stunnel, so it is encrypted.  The CPU may simply not be able to handle a lot of crypto load.

Remember in normal operation the CPU only needs to send 1Kb/s of to the cloud for management purposes.  Why bother putting in a much more expensive and faster CPU into a low end switch when the only benefit would be better speed test results run from the switch itself.

[MerakiDave]

@svetlinski The throughput test under any of the live tools pages is just measuring back-end mtunnel throughput for the control channel between your Meraki device and the Dashboard shard it's connected to back in the data center (cloud).  You can see what the speed test is actually doing (nothing exciting) by running a pcap and looking at all the TCP/7752 traffic.  It's not meant to be an indication of throughput for your clients to/from the Internet, and not necessarily from that actual piece of infrastructure to the Internet.  So for clients simply use your favorite & regular speed test tools for that. 

There are multiple factors that impact the test results when running a throughput test in the Dashboard, some of which are external/unrelated to your environment. 

More on that here:

https://documentation.meraki.com/zGeneral_Administration/Tools_and_Troubleshooting/Throughput_test_t...

[BlakeRichardson]

@svetlinski  I misunderstood what you were asking I thought you were meaning bandwidth used by port not the throughput test as @MerakiDave  mentioned. 

[BrechtSchamp]

I have experienced exactly the same issue and got exactly the same response from the helpdesk.

I extensively tested and I noticed that both my MX and my MR showed a realistic value for the throughput. The MS was indeed much lower. I found this weird as the AP has to go through the switch.

@PhilipDAth might be right and it might have to do with the cpu in the MS220-8P.

In any case I found the answer from helpdesk unsatisfying. It's called a throughput test so it should be testing throughput imo. If it's supposed to test connectivity it should be called that...

[svetlinski]

@BrechtSchampI know, right! 🙂 I think if they had provided a better answer, such as @MerakiDave 's reply above, I would have been more satisfied, wouldn't you?

@PhilipDAth, @BlakeRichardson  Thank you, I see what you mean now 🙂

I guess we'll just have to accept it for what it is and live with it, ay? 😉 

[Nick]

Thinking about it - I would agree its a hardware based limitation. The MS isn't geared to provide such high throughout. But as its just testing the tunnel to Meraki it doesn't need to be able to push great bandwidth so both are acceptable

[Mish]

The switching capacity of the MS220-8P is 20Gbps👈

https://meraki.cisco.com/lib/pdf/meraki_datasheet_ms220_compact.pdf

So it can definitely push way more traffic!

You guys are right, the "throughput" to the cloud test is labelled wrong. It's definitely NOT the MAX throughput. 9-10Mbps makes it look like a bottleneck when it definitely isn't.

I donno what the heck the engineers did to make the calculated speed always around 10Mbps!

The MX throughput is closer to a speedtest.net test.

[cmr]

It is to do with the CPU/code as stated previously, just because a switch can switch more than 1Gb of traffic, it doesn't mean it can generate it or receive it.  If you've ever had to tftp a new firmware to a switch it goes in kBps not mBps. Switching traffic is done in custom hardware that is designed purely to perform that function.  Because APs have to do a lot of different tasks they tend to have generic CPUs that are pretty powerful.  You can tell this by holding an AP that has been on for a while, MR32s in particular get pretty hot!

I tried on my home network and an MR55 gives pretty much the exact circuit throughout whereas an MS220 gives 10Mb.  At one of my work sites an MS225 gets to 14Mb which, to me, proves it is a cpu/code limited test.