Re: [AFMUG] speed tests and TCP/IP overhead

[dave]

exactly +1
 Also, it will compensate for the loss in some off the shelf routers 
when navigating the poorly written firewalls.



On 9/29/20 9:40 AM, Darin Steffl wrote:
On a 25 Mbps plan for example, we provision the speed bucket to 28 
Mbps flat. We use Preseem for the rate limiting. Generally on a 
speedtest, we see 26-27 Mbps on tests when provisioned for 28 Mbps on 
a 25 Mbps plan. If you provisioned to 26 Mbps, you would probably see 
exactly 25 Mbps.


I think it's great to overprovision to keep customers happy. The 
difference is negligible when you bump it 1-3 Mbps over the rate plan 
just to help the speedtest get at least what they pay for. It's easier 
than arguing with the customer.


On Tue, Sep 29, 2020 at 9:29 AM Ken Hohhof > wrote:


Two part question:

1) If you set a speed limit, like 10 Mbps, using Preseem or
Mikrotik queue or something similar, what do you expect to see at
a speedtest site like speedtest.net ?  I’m
thinking something like 9.5 Mbps.  I’m assuming the bandwidth
manager is looking at line rate but the speed test looks at payload.

2)  Do you “gross up” your speed limits so that the customer will
see 100% of the advertised speed when they run a speedtest?  If
so, by how much?  Or do you assume any reasonable person when they
see 9.5 Mbps but are paying for 10 will say “close enough”?  I
seem to remember that back when we were doing lineshared DSL,
Verizon grossed up the modem line rates, but AT&T didn’t.

-- 
AF mailing list

AF@af.afmug.com 
http://af.afmug.com/mailman/listinfo/af_af.afmug.com



--
Darin Steffl
Minnesota WiFi
www.mnwifi.com 
507-634-WiFi
Like us on Facebook 



-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] speed tests and TCP/IP overhead

[chuck]

One time I checked the difference was 6%

-Original Message- 
From: Matt Hoppes

Sent: Tuesday, September 29, 2020 1:44 PM
To: AnimalFarm Microwave Users Group ; Ken Hohhof
Subject: Re: [AFMUG] speed tests and TCP/IP overhead

We usually provision at 10% over the speed.

On 9/29/20 10:29 AM, Ken Hohhof wrote:

Two part question:

1) If you set a speed limit, like 10 Mbps, using Preseem or Mikrotik queue 
or something similar, what do you expect to see at a speedtest site like 
speedtest.net?  I’m thinking something like 9.5 Mbps.  I’m assuming the 
bandwidth manager is looking at line rate but the speed test looks at 
payload.


2)  Do you “gross up” your speed limits so that the customer will see 100% 
of the advertised speed when they run a speedtest?  If so, by how much? 
Or do you assume any reasonable person when they see 9.5 Mbps but are 
paying for 10 will say “close enough”?  I seem to remember that back when 
we were doing lineshared DSL, Verizon grossed up the modem line rates, but 
AT&T didn’t.





--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com 



--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] speed tests and TCP/IP overhead

[Matt Hoppes]

We usually provision at 10% over the speed.

On 9/29/20 10:29 AM, Ken Hohhof wrote:

Two part question:

1) If you set a speed limit, like 10 Mbps, using Preseem or Mikrotik 
queue or something similar, what do you expect to see at a speedtest 
site like speedtest.net?  I’m thinking something like 9.5 Mbps.  I’m 
assuming the bandwidth manager is looking at line rate but the speed 
test looks at payload.


2)  Do you “gross up” your speed limits so that the customer will see 
100% of the advertised speed when they run a speedtest?  If so, by how 
much?  Or do you assume any reasonable person when they see 9.5 Mbps but 
are paying for 10 will say “close enough”?  I seem to remember that back 
when we were doing lineshared DSL, Verizon grossed up the modem line 
rates, but AT&T didn’t.





--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] speed tests and TCP/IP overhead

[Nate Burke]

I think the big guys do that Too.  I remember getting a 100mb Comcast 
coax line, and it testing at 110mb,so the tech could point at his laptop 
and say O look!


I set the queue about 1-2mb higher than the plan rate.

On 9/29/2020 9:48 AM, Darin Steffl wrote:
Here are a couple screenshots showing my home at different speed 
buckets using Preseem.


First one is provisioned for 5.3 Mbps down, 1.5 Mbps up. I made sure 
to have no downstream traffic running but I do have one nest camera 
uploading about 300-400 Kbps.


It looks like on this first one, I lost very little from provisioned 
speed on a test. 5.07 Mbps to the set 5.3 Mbps rate.



Second speedtest is provisioned for 16 Mbps down, 6 Mbps up. Actual 
test shows about 15.2 down, 5.2 up.


So you lose a little going through the network and my tests were run 
on WiFi, not hardwired. They might show a little better if hardwired.




On Tue, Sep 29, 2020 at 9:40 AM Darin Steffl > wrote:


On a 25 Mbps plan for example, we provision the speed bucket to 28
Mbps flat. We use Preseem for the rate limiting. Generally on a
speedtest, we see 26-27 Mbps on tests when provisioned for 28 Mbps
on a 25 Mbps plan. If you provisioned to 26 Mbps, you would
probably see exactly 25 Mbps.

I think it's great to overprovision to keep customers happy. The
difference is negligible when you bump it 1-3 Mbps over the rate
plan just to help the speedtest get at least what they pay for.
It's easier than arguing with the customer.

On Tue, Sep 29, 2020 at 9:29 AM Ken Hohhof mailto:af...@kwisp.com>> wrote:

Two part question:

1) If you set a speed limit, like 10 Mbps, using Preseem or
Mikrotik queue or something similar, what do you expect to see
at a speedtest site like speedtest.net ?
I’m thinking something like 9.5 Mbps.  I’m assuming the
bandwidth manager is looking at line rate but the speed test
looks at payload.

2)  Do you “gross up” your speed limits so that the customer
will see 100% of the advertised speed when they run a
speedtest?  If so, by how much?  Or do you assume any
reasonable person when they see 9.5 Mbps but are paying for 10
will say “close enough”?  I seem to remember that back when we
were doing lineshared DSL, Verizon grossed up the modem line
rates, but AT&T didn’t.

-- 
AF mailing list

AF@af.afmug.com 
http://af.afmug.com/mailman/listinfo/af_af.afmug.com



-- 
Darin Steffl

Minnesota WiFi
www.mnwifi.com 
507-634-WiFi
Like us on Facebook 



--
Darin Steffl
Minnesota WiFi
www.mnwifi.com 
507-634-WiFi
Like us on Facebook 



-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] speed tests and TCP/IP overhead

[can...@believewireless.net]

TCP/IP overhead is 7% so we just add 10%. 10Mbps would be 11Mbps.

On Tue, Sep 29, 2020 at 10:41 AM Darin Steffl 
wrote:

> On a 25 Mbps plan for example, we provision the speed bucket to 28 Mbps
> flat. We use Preseem for the rate limiting. Generally on a speedtest, we
> see 26-27 Mbps on tests when provisioned for 28 Mbps on a 25 Mbps plan. If
> you provisioned to 26 Mbps, you would probably see exactly 25 Mbps.
>
> I think it's great to overprovision to keep customers happy. The
> difference is negligible when you bump it 1-3 Mbps over the rate plan just
> to help the speedtest get at least what they pay for. It's easier than
> arguing with the customer.
>
> On Tue, Sep 29, 2020 at 9:29 AM Ken Hohhof  wrote:
>
>> Two part question:
>>
>>
>>
>> 1) If you set a speed limit, like 10 Mbps, using Preseem or Mikrotik
>> queue or something similar, what do you expect to see at a speedtest site
>> like speedtest.net?  I’m thinking something like 9.5 Mbps.  I’m assuming
>> the bandwidth manager is looking at line rate but the speed test looks at
>> payload.
>>
>>
>>
>> 2)  Do you “gross up” your speed limits so that the customer will see
>> 100% of the advertised speed when they run a speedtest?  If so, by how
>> much?  Or do you assume any reasonable person when they see 9.5 Mbps but
>> are paying for 10 will say “close enough”?  I seem to remember that back
>> when we were doing lineshared DSL, Verizon grossed up the modem line rates,
>> but AT&T didn’t.
>> --
>> AF mailing list
>> AF@af.afmug.com
>> http://af.afmug.com/mailman/listinfo/af_af.afmug.com
>>
>
>
> --
> Darin Steffl
> Minnesota WiFi
> www.mnwifi.com
> 507-634-WiFi
> Like us on Facebook 
> --
> AF mailing list
> AF@af.afmug.com
> http://af.afmug.com/mailman/listinfo/af_af.afmug.com
>
-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] speed tests and TCP/IP overhead

[Adam Moffett]

Overhead depends on packet size and who's exactly measuring at what at 
what layer, but probably about 5-10%.  I bump customers up 10% so on a 
10meg they'll see something like 10.1 or 10.2.  I know Verizon FiOS does 
this too.  I figure if it saves even a handful of phone calls per year 
then it was worth it.



On 9/29/2020 10:29 AM, Ken Hohhof wrote:


Two part question:

1) If you set a speed limit, like 10 Mbps, using Preseem or Mikrotik 
queue or something similar, what do you expect to see at a speedtest 
site like speedtest.net?  I’m thinking something like 9.5 Mbps.  I’m 
assuming the bandwidth manager is looking at line rate but the speed 
test looks at payload.


2)  Do you “gross up” your speed limits so that the customer will see 
100% of the advertised speed when they run a speedtest?  If so, by how 
much?  Or do you assume any reasonable person when they see 9.5 Mbps 
but are paying for 10 will say “close enough”?  I seem to remember 
that back when we were doing lineshared DSL, Verizon grossed up the 
modem line rates, but AT&T didn’t.



-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] speed tests and TCP/IP overhead

[Darin Steffl]

On a 25 Mbps plan for example, we provision the speed bucket to 28 Mbps
flat. We use Preseem for the rate limiting. Generally on a speedtest, we
see 26-27 Mbps on tests when provisioned for 28 Mbps on a 25 Mbps plan. If
you provisioned to 26 Mbps, you would probably see exactly 25 Mbps.

I think it's great to overprovision to keep customers happy. The difference
is negligible when you bump it 1-3 Mbps over the rate plan just to help the
speedtest get at least what they pay for. It's easier than arguing with the
customer.

On Tue, Sep 29, 2020 at 9:29 AM Ken Hohhof  wrote:

> Two part question:
>
>
>
> 1) If you set a speed limit, like 10 Mbps, using Preseem or Mikrotik queue
> or something similar, what do you expect to see at a speedtest site like
> speedtest.net?  I’m thinking something like 9.5 Mbps.  I’m assuming the
> bandwidth manager is looking at line rate but the speed test looks at
> payload.
>
>
>
> 2)  Do you “gross up” your speed limits so that the customer will see 100%
> of the advertised speed when they run a speedtest?  If so, by how much?  Or
> do you assume any reasonable person when they see 9.5 Mbps but are paying
> for 10 will say “close enough”?  I seem to remember that back when we were
> doing lineshared DSL, Verizon grossed up the modem line rates, but AT&T
> didn’t.
> --
> AF mailing list
> AF@af.afmug.com
> http://af.afmug.com/mailman/listinfo/af_af.afmug.com
>


-- 
Darin Steffl
Minnesota WiFi
www.mnwifi.com
507-634-WiFi
Like us on Facebook 
-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

[AFMUG] speed tests and TCP/IP overhead

[Ken Hohhof]

Two part question:

 

1) If you set a speed limit, like 10 Mbps, using Preseem or Mikrotik queue
or something similar, what do you expect to see at a speedtest site like
speedtest.net?  I'm thinking something like 9.5 Mbps.  I'm assuming the
bandwidth manager is looking at line rate but the speed test looks at
payload.

 

2)  Do you "gross up" your speed limits so that the customer will see 100%
of the advertised speed when they run a speedtest?  If so, by how much?  Or
do you assume any reasonable person when they see 9.5 Mbps but are paying
for 10 will say "close enough"?  I seem to remember that back when we were
doing lineshared DSL, Verizon grossed up the modem line rates, but AT&T
didn't.

-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com