ISP speed
Paul Fuchs
tower broadcast. It is not very good but the only store in town. There
are no telephone lines where I live. Most people use cell phones.
Anyway, my question is about the ISP speed. My provider uses
speedtest.net to test the speed it is providing because it is
international iand is testing me, In theory, from a ISP in Costa Rica,
less than 50 miles away.
So my question is this. My last test gave me:
ping: 254 ms
download: 0.57 Mb/s
upload: 0.20 Mb/s
When I download a file from a USA site in Safari, I am getting speeds
against this reading of about 25 kBytes per sec. Now, by the rule of 8,
that works out to about 0.20kbps. So what's wrong with this picture?
--
During times of universal deceit, telling the truth
becomes a revolutionary act.
George Orwell
Ian Gregory
> When I download a file from a USA site in Safari, I am getting speeds
> against this reading of about 25 kBytes per sec. Now, by the rule of
> 8, that works out to about 0.20kbps. So what's wrong with this
> picture?
I think of it as a rule of 10. Yes there are 8 bits in a byte but I
figure there is also protocol overhead. If there is any packet loss
some packets will have to be delivered twice, adding to that overhead. I
would be surpried if what you are seeing is unusual.
How consistent is the 25 kBytes per sec figure? How often have you
measured it and was it at varying times of day etc?
Ian
--
Ian Gregory
http://www.zenatode.org.uk/ian/
Wes Groleau
> I think of it as a rule of 10. Yes there are 8 bits in a byte but I
> figure there is also protocol overhead. If there is any packet loss
Eight bits, plus a synchronizing bit at each end.
--
Wes Groleau
Fossilization … to teachers.
http://Ideas.Lang-Learn.us/barrett?itemid=1476
Richard Maine
> Ian Gregory wrote:
> > I think of it as a rule of 10. Yes there are 8 bits in a byte but I
> > figure there is also protocol overhead. If there is any packet loss
>
> Eight bits, plus a synchronizing bit at each end.
That's for serial protocols (i.e. dialup). The overhead for TCP/IP is
usually much lower. Well, it can be much higher for things like
interactive stuff where you might actually often send a packet with only
one data byte, but then throughput isn't typically the most important
spec for that kind of application.
I recall being pleasantly surprised at my first "broadband" connection
(a 128k IDSL line, as I was and still am too far from the CO for
"regular" DSL, though I'm using cable instead now). I measured the
throughput and got what seemed to me like better than the advertized
speed. Then I realized that I had been dividing by 10 out of habit from
serial connections, but that 8 was more appropriate for TCP/IP. (Sure,
it isn't exactly 8, as there is some overhead, but close enough). When I
used 8, I found it was pretty much right on the advertized speed.
--
Richard Maine | Good judgment comes from experience;
email: last name at domain . net | experience comes from bad judgment.
domain: summertriangle | -- Mark Twain
Wes Groleau
> Wes Groleau <Grolea...@FreeShell.org> wrote:
>> Eight bits, plus a synchronizing bit at each end.
>
> That's for serial protocols (i.e. dialup). The overhead for TCP/IP is
IP is at a higher level than the start/stop bits.
TCP is higher still.
DSL, Verizon FIOS, Cable, and ethernet do use serial protocols.
However, I reckon you're right that they don't need a synchronizing
bit on every octet.
> I recall being pleasantly surprised at my first "broadband" connection
> (a 128k IDSL line, as I was and still am too far from the CO for
> "regular" DSL, though I'm using cable instead now). I measured the
> throughput and got what seemed to me like better than the advertized
> speed. Then I realized that I had been dividing by 10 out of habit from
> serial connections, but that 8 was more appropriate for TCP/IP. (Sure,
> it isn't exactly 8, as there is some overhead, but close enough). When I
> used 8, I found it was pretty much right on the advertized speed.
Mine seems to keep bytes per second about a tenth of bits per second.
--
Wes Groleau
Standards?a parable
http://Ideas.Lang-Learn.us/WWW?itemid=145
E Z Peaces
> I am out in the boonies of Costa Rica. I managed to get internet via
> tower broadcast. It is not very good but the only store in town. There
> are no telephone lines where I live. Most people use cell phones.
>
> Anyway, my question is about the ISP speed. My provider uses
> speedtest.net to test the speed it is providing because it is
> international iand is testing me, In theory, from a ISP in Costa Rica,
> less than 50 miles away.
>
> So my question is this. My last test gave me:
> ping: 254 ms
> download: 0.57 Mb/s
> upload: 0.20 Mb/s
>
> When I download a file from a USA site in Safari, I am getting speeds
> against this reading of about 25 kBytes per sec. Now, by the rule of 8,
> that works out to about 0.20kbps. So what's wrong with this picture?
>
I don't know if it's related, but the other day I tried BitTorrent to
download NeoOffice. The file came in at a much slower speed than the
network traffic shown in Activity Monitor. I've also heard that files
download more slowly if you use a proxy server.
Paul Fuchs
> How consistent is the 25 kBytes per sec figure? How often have you
> measured it and was it at varying times of day etc?
It will vary. Lately it will not go over 40 and falls down into the
teens quite often. But 25 is a typical average. It tends to be slower
in the evening, presumably because higher usage is clogging the system.
Speedtest also gives me a ping number, and this varies tremendously. I
have seen it run from 40 ms to 1300ms. The Safari page load speeds seem
to be more related to the ping speed than variations in the download
speed. This morning the system seemed quite spiffy, and when I checked
it, the download was typical but the ping was 51, which is about as fast
as it gets.
I also suspect that the slow upload speeds slow the system a lot because
it has to wait longer for the error checking. Typical download is 500
kbps and upload 100. Could this be a major reason why the file download
speed, even using the rule of ten, seems a lot slower than the speedtest
download? The rule of ten should give me at least 50 kBytes/s and I
never get close to this.
Kevin McMurtrie
pf@porkain'tkosher.oink (Paul Fuchs) wrote:
> I am out in the boonies of Costa Rica. I managed to get internet via
> tower broadcast. It is not very good but the only store in town. There
> are no telephone lines where I live. Most people use cell phones.
>
> Anyway, my question is about the ISP speed. My provider uses
> speedtest.net to test the speed it is providing because it is
> international iand is testing me, In theory, from a ISP in Costa Rica,
> less than 50 miles away.
>
> So my question is this. My last test gave me:
> ping: 254 ms
> download: 0.57 Mb/s
> upload: 0.20 Mb/s
>
> When I download a file from a USA site in Safari, I am getting speeds
> against this reading of about 25 kBytes per sec. Now, by the rule of 8,
> that works out to about 0.20kbps. So what's wrong with this picture?
It could be that you're suffering from packet loss due to malfunctioning
equipment or heavily loaded networks sending traffic to you in bursts.
Try pinging a site as you're downloading. For Mac OS 10.5 and earlier,
packet loss has an extremely high performance penalty on high latency
connections. Every lost packet causes the stream in progress to be
discarded, repositioned, and then resumed. That takes a while when
there's 1/4 second of lag. Mac OS 10.6 finally implements "Selective
ACKs" which allows lost packets to be resent without interrupting the
stream in progress.
--
I won't see Google Groups replies because I must filter them as spam
Jean-Michel LEON-FOUN-LIN
> When I download a file from a USA site in Safari, I am getting speeds
> against this reading of about 25 kBytes per sec. Now, by the rule of 8,
> that works out to about 0.20kbps. So what's wrong with this picture?
Is it not 0.2 MEGA bits per second (not 0.2 KILO bits per sec.)?
Paul Fuchs
Yes, it should have read 0.2 Mbps. I didn't correct it with a follow
post, because the error was obvious from the other information.