Manual setting for Internet speed as Speedtest is unreliable.
Manual setting for Internet speed as Speedtest is unreliable.
Hi All,
Stil the speedtest is unreliable. I have a 1gb/1gb line where I also REALLY get 940/940.
But the speedtest gives me 538/538 max which is incorrect.
I asked already since day 1 to correct this but it is still not done.
Can you please take a look and fix the error or at least let ME set the linespeed.
- Copy Link
- Subscribe
- Bookmark
- Report Inappropriate Content
Dear All,
We are now aware of the issues mentioned and have fed back to the TP-Link team.
Further research on this issue is on the way.
Thanks for your all feedback.
Best Regards!
- Copy Link
- Report Inappropriate Content
Likewise, I first blamed the internet-provider as during migrating to a fully tp-link network to GB network I measured using the every 10 hour feature significant lower speeds.
After the required testing with a PC directly attached to the provider modem the provided internetspeed was conform spec.
Also testing, using the direct attched pc to the mode, at different times during the day and week all measurements are witin spec.
Again I used the WAN speedtest of the ER7206 (using the OC200 omada interface) and ensuring the test was the only wan usage (clients and server down) the numbers are still way off 300Mb/s up to 710Mb/s but never the 950Mb +50Mb was measured. According to TP-Link provided spec the ER7206 should be able to handle 950Mb/s NAT load so I expected to simular max WAN speed test results. The CPU and mem usage during the test was minimal so of no impact.
My conclusion the speedtest performed by the ER7206 (latest firmware level) is a rendom number generator and has noting to do with speed max testing of the provider network.
Questions:
- In this forum I did not see a lot of complains about the speedtest, did I misuse or is it my mis interpretation of this feature?
- Are there more simular findings or is this finding tipical for high bandwidth networks?
- Copy Link
- Report Inappropriate Content
@TheMuk using several er605 and getting everywhere full speed, on several locations I've set in omada to run speedtest every 24 hours, using 500/500Mbit (coax cable) on wan, 300/50Mbit (coax cable) on wan1, 250/25Mbit (lte, 5g) on wan2 and wan3 with 150/25 (lte, 4g) but I do not have 1gbit like you. I read that some complain argumenting speedtest shows it higher, where other report that it does give full gbit, I cant confirm nor deny for 1Gbit, but 500Mbit connections with er605 delivered always ~500Mbit for wan, always 300/50 for wan1, just lte connections vary but only due to all other factors just not my network (using external outdoor antennas). So far I saw no issue with er605 on locations where I used it and my mobile application shows exactly the speed I get on my dashboard, where omada does use iperf for speedtest and I would consider those tests more accurate than speedtest as long as you do not know if your ISP throttles your connection.
Not sure if it helps, but here would be few tips in general:
I. look at your cpu utilization as well as how many clients it has, would show if it is overloaded, especially if not overloaded, it should be able to reach 1gbit.
II. Speedtest is simply not reliable for exact testing, but iperf is (iperf3 is my preference).
III. It is also well known that some ISP's use dirty tricks in actually disabling throtling for specific service, one of their favourite service for that is speedtest.net by ookla, where there are many sites acting as speedtest but use ookla servers. In some countries some ISP's which are not rogue want to add unlimited bandwidht for netflix/prime and so on... . So far, you can not 100% rely on speedtest, the only real testing is if you do test it with your devices, eap's come with iperf which you can run, as example, after you put cables in your house, despite that some cheap lan tester says that cable is ok, you want to see if two devices connected with that cable can reach their interface speed, lets say 1Gbit. Following this in general is what I do to check performance once I've setup cables, network dose's, switches, routers, etc..:
1. testing beetwen two pc's in your network shows quickly if you reach 1Gbit, if lan ports/your cards work, if your switch is able to deliver 1Gbit on both ports and if cable is giving that speed. reaching beetwen 900-1000Mbit/s is normal for gigabit LAN's.
2. After you have tested your local network, test your eap's, iperf is preinstalled, just ssh and and run iperf -s, on your pc connected over lan (or binding lan interface), run iperf -c IPOFYOUREAP -p 5001 -P 10, then the same with wireless, eap225's as example do not really deliver with iperf more than 200-300, but that has other reasons, however, by that you will know how your eap's perform with lan and wireless.
3. On current point if lan delivers prommised speeds you know exactly that everything works, as next test your wan with your pc, depending on where you are located, best is always to take closest to you server, here are some from EU including 10Gbit+ : https://iperf.fr/iperf-servers.php , this one should be 10Gbit: https://speedtest.wtnet.de/ . There are more iperf servers, one can also get own vps with 1/10Gbit and run iperf on it. Rogue ISP's could disable throtling for all publicly known servers, meaning that having own vps with 1/10Gbit can be used.
4. iperf test of vpn
Those results deliver more exact results, as well as one can modify a lot preferences for iperf too. With iperf3 use -V switch which will show extended statistics like cpu usage.
Also, if you use multiple wan's, then run iperf on each wan to see how those wan's perform, there you will see if there are any bottlenecks or if something is missconfigure, as example if I have 1 Gbit line, 100Mbit line, 1Mbit line, then I for sure do not want to use loadbalancer, 100Mbit and 1Mbit can act as backup if 1Gbit fails as well for device which love to phone home.
Considering you have done all those tests and you are 100% sure that without omada router you get full gbit and with omada router just 300-500. Consider then doing another round of iperf tests from your pc without omada router to see if you would achieve the same speed of speedtest which is 1Gbit, then it would be clear that your ISP does not throttle your connection and your device might be faulty, hope support can help you out in that case.
If you did not do iperf tests, then I encourage you to do it, best writting simply a script, as that is the most comfortable way to find out how much your interface delivers and how your network really performs, independently from throttling if there is any.
If you find out your providers throttles you, best is to change provider if there is any offering the product you need, if not, well, then you have no luck and probably with vpn you can void bad behaviour of your ISP.
- Copy Link
- Report Inappropriate Content
- Copy Link
- Report Inappropriate Content
@TheMuk I believe you and at no point I wanted to put your experience in question, as my fastest is 500 and that is delivered, I even can't reproduce your issue because I would need faster connection than I have, just wanted to give some hints if it helps, as long as you get the speed statistic is not important, guess support can say more if iperf is in some way restricted or why it happens, but it would annoy me too if it is buggy, maybe it is a software bug, did you reproduce same results with lower controller/firmware versions, do you remember?
- Copy Link
- Report Inappropriate Content
- Copy Link
- Report Inappropriate Content
- Copy Link
- Report Inappropriate Content
@jandico you are welcome. I guess it is a software bug as neither your router is overloaded nor your isp throttles it, thats clear. I hope it is a software bug, as then it is probably quickly resolvable. Also, you use more powerfull router than er605, er605 delivers 1gbit on wan ports.
If you really want to set max output of your lan ports, than connect it simply on wan ports to your existing network which is 1gbit. That way you will know very exactly what the max performance you would achieve, then you can be at least secure how much ports deliver.
Also, run iperf also from multiple pc's within local network to get real max output of all ports.
I am also not sure if you are familiar with bonding as well which mode would fit best, especially if you want to test all 48 ports of a 48 port switch :) with few device and multiple usb gigabit adapters, as example you can combine lan and wireless on a notebook too test not only your local switches but also how your eap's perform when your switch is overloaded. Check out debian wiki for bonding: https://wiki.debian.org/Bonding
It sounds complicated, but read also projects readme's there is more explanation about modes and parameters and in real its quite simple.
As example, on one of my industrial case pcs which I kept in a company, has 6 gigabit ports on it which was perfect for bonding. I do not believe that bonding would resolve your issue, why I mention bonding is because you do have 1Gbit internet as well as your local network is also 1Gbit, you kinda run into a bottleneck if you move files and run internet on full speed, for that you might want to get at least one additional gbit port to have 2gbit. There are also switches supporting different modes like 802.3ad, link aggregation etc.., but with bonding you could use it without checking for your switches compatibilty of specific modes (which you might not need).
If you have 10Gbit network then you can ignore my suggestion about bonding, as 10Gbit has enough ressources if you internet is used in full swing.
EDIT: typo again, I mean of course 802.3ad with switch supporting it, not 802.11ad.
- Copy Link
- Report Inappropriate Content
My guess is that the speed test requires flash storage or something to run and its too slow on the tp-link hardware.
- Copy Link
- Report Inappropriate Content
@mackworth I hope not, I did not really inspect which way it is checked, but eap's have iperf on them, where er605 as example has no iperf binary, I would rather believe that speedtest is feature of omada cloud controller which does use iperf which is done then with controllers port. Beside that, there is enough space in ram for those test which do not really require free space, I would exclude it, but as I have no clue, you might be right it, however, I doubt speedtest is performed using additional space, especially as eap devices have read only system, if you want to flash as example openwrt, you can use free ram space in tmp folder for that.
Also, interesting to compare iperf and kernel where I must admit, I am surprised it is such old kernel on latest fw version. Here one would wish ability to build own builds. Also iperf3, simply compile it for mips, copy to ram on eap and run from there. So, all in all one can use third party binaries on a eap's, its just that those are gone after reboot.
- eap225outdoor(EU) v1, 5.0.7: iperf version 2.0.5 (08 Jul 2010) pthreads
Kernel: Linux EAP225-Outdoor 3.3.8 #1 Tue Oct 26 08:58:44 CST 2021 mips GNU/Linux
- eap235-wall(EU) v1, 3.1.0: iperf version 2.0.5 (08 Jul 2010) pthreads
Kernel: Linux EAP235-Wall 3.10.108 #1 SMP Wed Jul 21 12:40:40 CST 2021 mips GNU/Linux
- eap245(EU) v3.0, 5.0.4: iperf version 2.0.5 (08 Jul 2010) single threaded
Kernel: Linux EAP245 3.3.8 #1 Thu Oct 21 15:35:35 CST 2021 mips GNU/Linux
We see here that different binaries are used, but no matter if I run iperf on them as a server (download) or client (upload) over lan or wireless, its always around 226Mbit, where that specific device reaches easily 250Mbit over 5Ghz (that is its wan speed). If iperf v2 is used for speedtest values on router, then this as example could be issue. I had once one version of iperf3 which in local network simply refused to go over 500, but only in server, in client mode it was 1Gb, few days later there was update and issue was gone. This sounds very similar, but I must admit I did not inspect so far how omada performs speedtest, I just assume iperf could be involved, and if not, it is a good question why not.
@fae which kernel is with current eap610-outdoor firmware, hope it is not v3?
- Copy Link
- Report Inappropriate Content
Dear All,
We are now aware of the issues mentioned and have fed back to the TP-Link team.
Further research on this issue is on the way.
Thanks for your all feedback.
Best Regards!
- Copy Link
- Report Inappropriate Content
Information
Helpful: 2
Views: 2907
Replies: 16