Kitz Forum

Internet => Interesting Websites => Topic started by: Weaver on March 04, 2019, 06:59:27 AM

Title: Thinkbroadband speedtester IPv6 vs IPv4 mode
Post by: Weaver on March 04, 2019, 06:59:27 AM
Iíve been trying to work out whether the supposed poor upload performance reported by some speedtester is totally bogus or not. The TBB speedtester (https://www.thinkbroadband.com/speedtest) gives me 1.0 Mbps IPv4 upload but 1.4-1.6 Mbps IPv6. Perhaps the tester is less busy in IPv6 mode, but no, itís 0640, so there canít be anyone about surely? I do have things set up so that IPv6 packets are only 1408 bytes max, and IPv4 would be 1500, so perhaps that might conceivably affect the server?

Does anyone else get an IPv6 vs IPv4 difference from this?
Title: Re: Thinkbroadband speedtester IPv6 vs IPv4 mode
Post by: jelv on March 04, 2019, 09:27:10 AM
I've just looked back through my result history. If anything I see very slightly higher results from IPv4 - but it is marginal and on some tests run in quick succession it is the other way round.

IPv6
(https://www.thinkbroadband.com/_assets/speedtest/button/1541771320877271155.png) (https://www.thinkbroadband.com/speedtest/results.html?test=1541771320877271155)

IPv4
(https://www.thinkbroadband.com/_assets/speedtest/button/1541771370200427255.png) (https://www.thinkbroadband.com/speedtest/results.html?test=1541771370200427255)

I'd suspect something to do with the way the bonding works - perhaps it is more efficient in IPv6.
Title: Re: Thinkbroadband speedtester IPv6 vs IPv4 mode
Post by: Weaver on March 04, 2019, 09:07:39 PM
From the evidence of that thinkbroadband website, my upload performance efficiency problem has improved. I was moaning that I wasnít getting anything like the upload speed expected from calculating it pretty accurately in the sense of deducting absolutely all overheads properly. Also I seemed to remember that a year or so upload speeds were much better.

I was wondering if the poor results were due to incorrect load-share upstream figures per link set in my router, inaccurate because they were out of date for line 3 - the slow-upstream link - whose speed keeps changing. I now have a program that uploads the right numbers into the Firebrick routerís xml config for per-line upstream capabilities, so I ran that to ensure load shares were up-to-date and retested. Results were inconclusive because the TBB speed tester has too much variation in the numbers.

Mind you, that doesnít have anything to do with IPv6 vs IPv4 upstream of course.

I was moaning tonight because FaceTiming my wife who us on the same WLAN was so rubbish. <rant> The thing must be too stupid to route it properly and keep it inside the LAN I think. Iím assuming that the idiot thing really does go outside and back in, even though we all have global routable IPv6 addresses that clearly show we are in the same /64. And in the case of IPv4 you can tell as well. But in general, why not just ping (metaphorically) dest to try it? </rant>