Kitz Forum

Internet => General Internet => Topic started by: parkdale on November 08, 2018, 04:59:43 PM

Title: Speed test
Post by: parkdale on November 08, 2018, 04:59:43 PM
Has anyone else come across this speed test?
https://fast.com/#
Title: Re: Speed test
Post by: kitz on November 08, 2018, 05:48:28 PM
Not previously.

I quite like how it has configurable options such as single or multiple threads and duration of test.

Not quite sure how accurate it is, as both times I ran the test it said spot on 60Mbps which seems a convenient round figure.  TBB recorded 61.3.
Title: Re: Speed test
Post by: parkdale on November 08, 2018, 07:02:57 PM
Mine came to 46Mb dl and 17Mb ul 11ms so not too far away from my normal line speeds
Title: Re: Speed test
Post by: chenks on November 08, 2018, 07:05:58 PM
220Mbps down / 9.7Mbps up
seems reasonable although the up is reporting a little low.
Title: Re: Speed test
Post by: licquorice on November 08, 2018, 07:55:43 PM
 ???
Title: Re: Speed test
Post by: jelv on November 08, 2018, 10:04:07 PM
I set it to maximum of 6 parallel connections to match tbb. The results have varied from 64 to 68. TBBx1 usually gives me around that with TBBx6 usually 69 or 70.
Title: Re: Speed test
Post by: Weaver on November 08, 2018, 11:16:34 PM
With defaults it said 8.2 Mbps down, 1.3Mbps up which is believable since line 4 is very faulty now, only running at about half speed if set to 6dB. Line 4 went bad in 1 Nov and I have been very ill the past few days so did not do anything about it.

I set the tester to minimum of four, maximum of twelve parallel connections and it gave a crazy result of 21 MBps, and the upstream speed is 1.4 Mbps now. It said it was using IPv6.

That means that with those settings, it is by far the most broken tester I have ever seen.

I had a thought. I wonder if they have forgotten to multiply by the PDU size, as my IPv6 max size PDUs will be only 1408 bytes long now, not 1500. (This is because I reduced the MTU to be less than the IPv6 max MTU 1420 = 1440-20, which is set to be compatible with an IPv6 6in4 tunnel through a 3G dongleís IPv4 MTU 1440 during failover, that is the state when all four dsl lines fail. 1408 = n * 48 - 32, where PPPoEoA overhead is 32 bytes and 20 bytes is the 6in4 tunnel overhead. So 1408 bytes plus 32 bytes of overhead will be an exact number of filled ATM cells.)
Title: Re: Speed test
Post by: burakkucat on November 08, 2018, 11:31:53 PM
I do not trust any throughput speed tester.  :no:  I recommend choosing just one and then to only consider the delta between any two successive measurements.
Title: Re: Speed test
Post by: Weaver on November 08, 2018, 11:41:01 PM
I agree entirely with Burakkucat. Since they donít even say what it is that they are reporting then itís impossible to have any confidence. One can of course simply time a file download and then at least you know what you are measuring, but that doesnít give you the link speed, it gives you the result of a whole load of stuff involving the behaviour of say TCP as well.

I think someone should publish an RFC for a highly accurate speed tester methodology that is software and protocol independent (because the exact algorithm is given in the RFC). I would use a method of increasing the rate of timed packet sends, with packets of different sizes, while the receiver records the number of packets received at each speed and then the receiver and sender talk to one another and compare egress rates and ingress rates over time to determine a steady-state, no-loss rate or minimum loss rate.
Title: Re: Speed test
Post by: CarlT on November 09, 2018, 01:45:40 PM
Yes, OP, though not used it for a while. It was fine last time I did.

Weaver - I hope not. Ideally an RFC saying people should stop fixating over how fast their connection is and just use the thing. Some people's testing borders on obsessive which is bizarre.
Title: Re: Speed test
Post by: Weaver on November 09, 2018, 02:13:02 PM
Youíre absolutely right about obsessive testing, however speed is something that we pay money for and service providers and equipment manufacturers do sometimes need to be held to account so we need agreed reproducible and meaningful standards. Indeed some applications will simply fail if link speed is inadequate and some users therefore need solid guarantees as part of a service agreement,
Title: Re: Speed test
Post by: chenks on November 09, 2018, 02:13:48 PM
Youíre absolutely right about obsessive testing, however speed is something that we pay money for and service providers and equipment manufacturers do sometimes need to be held to account so we need agreed reproducible and meaningful standards.

the end-user does have to accept it's an "up to" service though.
Title: Re: Speed test
Post by: CarlT on November 09, 2018, 02:44:02 PM
Indeed some applications will simply fail if link speed is inadequate and some users therefore need solid guarantees as part of a service agreement,

Consumer-grade broadband is the wrong product for those that may need 'solid guarantees' as part of a service agreement.
Title: Re: Speed test
Post by: Weaver on November 09, 2018, 08:02:19 PM
Agreed. But I was talking about methodologies of testing. I do think there should be a public competition, beauty contest, to choose a respected and solid testing method.
Title: Re: Speed test
Post by: renluop on November 09, 2018, 08:49:48 PM
22 Mbps vs Speed Test Net 23.5 and TBB 23.3. Sync 25258 kbps
Title: Re: Speed test
Post by: CarlT on November 09, 2018, 08:54:49 PM
Already largely done.

https://tools.ietf.org/html/rfc6349
https://tools.ietf.org/html/rfc2544