ISP speed
in [Mac Systems]
|
Prev: Odd USB Extension Cable
Next: FrankenMac?
From: Paul Fuchs on 6 Jan 2010 17:28 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? -- During times of universal deceit, telling the truth becomes a revolutionary act. George Orwell
From: Ian Gregory on 6 Jan 2010 19:44 On 2010-01-06, Paul Fuchs <pf(a)porkain'tkosher.oink> wrote: > 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/
From: Wes Groleau on 6 Jan 2010 23:19 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. -- Wes Groleau Fossilization … to teachers. http://Ideas.Lang-Learn.us/barrett?itemid=1476
From: Richard Maine on 6 Jan 2010 23:34 Wes Groleau <Groleau+news(a)FreeShell.org> wrote: > 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
From: Wes Groleau on 6 Jan 2010 23:59
Richard Maine wrote: > Wes Groleau <Groleau+news(a)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 |