Internet?

[Electroenthusiast]

How do internet Providers Provide Internet?
i.e., how to they connect to all the servers in the world?
Is Satellite used?

 

[edeca]

You might like to read this page: HowStuffWorks "How Internet Infrastructure Works"

Yes, satellite is used for some portions of the internet. But it is relatively slow and therefore not as good for VOIP as undersea cables. Satellites are also expensive both to launch and maintain.

 

[Electroenthusiast]

You might like to read this page: HowStuffWorks "How Internet Infrastructure Works"

Yes, satellite is used for some portions of the internet. But it is relatively slow and therefore not as good for VOIP as undersea cables. Satellites are also expensive both to launch and maintain.

Hmmm... Late Reply
Where is the Domain name System? In US?? Each country has it's Own??
Why Internet Speed as what we get is Kbps(bits), then why do ISP define them in terms of KBps(bytes)?
They are just confusing.

 

[Gobbledok]

Hi Electroenthusiast,

I don't know much about DNS servers, but the likely reason that ISP's rate their speeds differently (KBps or Kbps) is marketing.

8000Kbps sounds faster than 1000KBps.

 

[Electroenthusiast]

Hi Electroenthusiast,

I don't know much about DNS servers, but the likely reason that ISP's rate their speeds differently (KBps or Kbps) is marketing.

8000Kbps sounds faster than 1000KBps.

Yes, Maybe... It's like misleading the customer.

 

[Nigel Goodwin]

Yes, Maybe... It's like misleading the customer.

Historically baud rates have always been that way - it's the only way that makes sense - bit rate is universal, byte rate depends on the size of the byte and the supporting bits used.

 

[Electroenthusiast]

Historically baud rates have always been that way - it's the only way that makes sense - bit rate is universal, byte rate depends on the size of the byte and the supporting bits used.

IMHO, 1 byte = 8 bits - that's a constant nowadays. I feel that the byte rate remains same and common throughout the world.

 

[edeca]

Historically baud rates have always been that way - it's the only way that makes sense - bit rate is universal, byte rate depends on the size of the byte and the supporting bits used.

Well I'd suggest that a byte is always a "byte" (one octet or 8 bits) in modern networking terms, but there is still an important point in here. There are two main considerations, a technical issue (about throughput) and a marketing issue (about how ISPs sell their services).

The technical issues are varied and include:

- There are lots of overheads in networking protocols (IP and TCP headers from your machine, repackaged with PPP and ATM headers from the DSL router to the ISP etc) which mean that the entire bandwidth is not exclusively used for "downloading data". Nigel refers to these as "supporting bits".

- Kilobits per second (10^3 or 1000 bits) don't equate to kibibits (2^10 or 1024 bits), the classic "network vs. storage" comparison. This makes a bigger difference if you are downloading a 1GiB file, at 1MByte/s it will take more than 1000 seconds.

- You will likely be contended on any domestic internet style connection, sharing the total capacity available between a number of people. Whilst this is less of an issue now than 5 years ago there are still obvious limits, an ISP with 1000 customers doesn't have 8gbit/s (8mbit/s * 1000) worth of transit. And most ISPs have tens of thousands of customers.

As for marketing, companies use all sorts of terms which make no sense to consumers. Subtle advertising suggesting that fibre is "better" than copper, phrases such as "up to 8Mbit/s" etc. Many customers have no idea that the "wireless router" they bought is only wireless in the house..

 

[Electroenthusiast]

....
- There are lots of overheads in networking protocols (IP and TCP headers from your machine, repackaged with PPP and ATM headers from the DSL router to the ISP etc) which mean that the entire bandwidth is not exclusively used for "downloading data". Nigel refers to these as "supporting bits".

I had never thought in this direction, thanks edeca. Thats why, the ISP tells that speed is 1024kbps(1Mbps), in that case we should get 128KBps download speed, but we get much lesser speed because of the above mentioned fact.

Kilobits per second (10^3 or 1000 bytes) don't equate to kibibits (2^10 or 1024 bytes), the classic "network vs. storage" comparison.

1 Kilo bits = 1024/8 Bytes..

As for marketing, companies use all sorts of terms which make no sense to consumers. Subtle advertising suggesting that fibre is "better" than copper, phrases such as "up to 8Mbit/s" etc. Many customers have no idea that the "wireless router" they bought is only wireless in the house..

Yes i know, some ppl have there max transfer rate of their wifi accesspnts to be like 1Mbps, even then they have buy a 16Mbps internet service.

And another ques, the speed keeps varing around some rate. Why this happens? How do they limit the speed based on customers tariff/package, and how they keep the count of bandwidth consumed?

I have a cable broadband, and people say that it's not LAN and not DSL. How are these terms important?

And, this is a faulty question, ;D ... One of my fren had asked me this, while we were discussing on something. He said: 'The speed of all electronic signal(including Internet) is equal to spd of light. Then, why does ISP decrease the speed to Mbps, Kbps and so on...'. What he meant was all these signals propagate at that high speed, why and how the speed gets decreased. And i did answer it, but not sure whether i was right.

Video:
https://www.youtube.com/v/Ve7_4ot-Dzs

 

[Nigel Goodwin]

IMHO, 1 byte = 8 bits - that's a constant nowadays. I feel that the byte rate remains same and common throughout the world.

It certainly doesn't - even if you consider a 'byte' is always 8 bits - the number of extra bits can vary. Commonly 10 bits are used for eight bit bytes, but 9 can be used, 11 can be used - and those are just the standard ones. Generally if you're converting BPS to BytesPS, you should divide by ten - unless you have info otherwise.

You also need to bear in mind 8 bits 'bytes' is a relatively 'recent' thing, octal was far more common before hexadecimal took over, and bit sizes have varied greatly other the years.

Considering a 'byte' is 8 bits, you couldn't even use it for speed in an octal based system (3 bits, 6 bits, 9 bits etc.)

 

[edeca]

1 Kilo bits = 1024/8..

No. Kilo as an SI prefix is 10^3, or 1000. This is the same in computing and networking.

My original message does mix bytes and bits though, blame that on lack of coffee. I'll edit so it's clear when you come back to it in 2 more years

 

[edeca]

It certainly doesn't - even if you consider a 'byte' is always 8 bits - the number of extra bits can vary

Whilst correct in general terms this is irrelevant for anything above the physical layer in networking (ManchesterII or 8b10b). It is also irrelevant for any calculation involving networking speeds involving bytes.

 

[Sceadwian]

edeca. The SI unit has nothing to do with binary notation.
1kilobyte is 1,024 bytes, 1 megabyte is 1024 kilobytes etc.. If the specific use of SI units has not been clearly made then the results should be in true binary notation, not SI units.

Hard drive makers have redefined these terms to mean 1000 bytes = 1kilobyte and 1000kilobytes = 1megabyte to artificially inflate their capacities. But they are explicitly defined in binary notation as being derived from multiplication of 1024. Read the back of the box/paper of any hard drive and you'll see the re-definition of binary notation clearly.

All of the bandwidth speed tests that are out there are measure of delivered data not raw bandwidth, the actual data transmitted is dependent on the packet size, but this is irrelevant seeing as how the data delivered is what should be measured anyways, not overhead. The proper term for this I believe is throughput, so basically the bulk of all speed tests sites out there are using the wrong terminology.

Very few to no one notices because all they care about is big numbers. The technicalities get swept to the wayside.

Personally I measure my practical bandwidth by the speed up/down of a large sized torrent, as that's a more practical value of what can be achieved in a practical situation... as well it's a practical situation =)

 

[edeca]

edeca. The SI unit has nothing to do with binary notation.

It has everything to do with modern computer networking, which is what this thread is talking about. It was standardised many years before I studied it.

Both of these tiresome pages from Wikipedia explain:

https://en.wikipedia.org/wiki/Data_rate_units
https://en.wikipedia.org/wiki/IEC_60027-2

Sure, 1 kilobyte is widely and incorrectly taken as 1024 bytes. But that certainly hasn't been the standard since 1999 (!) and all studying I did to degree level in networking used multiples of 1000 for throughput calculations. Note that the IEC amendment applies to electronics as well as telecommunications, I wasn't aware of that.

 

[Nigel Goodwin]

It has everything to do with modern computer networking, which is what this thread is talking about. It was standardised many years before I studied it.

Both of these tiresome pages from Wikipedia explain:

https://en.wikipedia.org/wiki/Data_rate_units
https://en.wikipedia.org/wiki/IEC_60027-2

Sure, 1 kilobyte is widely and incorrectly taken as 1024 bytes. But that certainly hasn't been the standard since 1999 (!) and all studying I did to degree level in networking used multiples of 1000 for throughput calculations. Note that the IEC amendment applies to electronics as well as telecommunications, I wasn't aware of that.

The Wiki you linked to also says:

"The standardized binary prefixes such as Ki- were relatively recently introduced and still face low adoption"

and

"Computer and technology industries have yet to adapt to these standards."

So it sounds like no one apart from you and your course pay any attention to them

Interesting though, I'd never heard of it before - usual stupid change for no reason!.

 

[edeca]

"The standardized binary prefixes such as Ki- were relatively recently introduced and still face low adoption"

Recently as in 1999! That's the fast paced change of technology for us all

 

[Sceadwian]

edeca, do you really think that people that have been using binary systems for most of their life are going to adopt a new silly slang language like kibi mebi gibi and tebi without resistance? Instead of just saying 'you're wrong' you should post a link to the appropriate site that shows how the notation should be.

I'll wager that most digital pundits weren't even aware that there was an SI prefix system adopted. The reason being that the only people that would even have a clue is those that actually create new standards. It's really sad if you think about it though, because if you look at all the PDF's for chips made from 1998 and on I'm pretty sure you'll find the use of the SI binary prefix virtually non-existent, even though the binary intentions are still fully valid..

Apparently someone forgot to tell the semiconductor industry =)

Anyone have a PDF for a micro controller or digital device that uses the SI binary pre-fix system?

I'd call that a major marketing fail for SI.

 

[KeepItSimpleStupid]

Back to the bits per second. I was exposed to it from the get-go with baud rates of 110 baud and 134.5 baud and up to 1200 baud which was "FAST" back then.

You has 1 start bit, 7 or 8 data bits, 0 or 1 parity bit and 1, 1.5 or 2 stop bits. Two stop bits were necessary so the motor had time to stop.

The usual 1 start, 8 data and 2 stop so baud/10 sometimes is characters per second.

The start/stop bits are Overhead.

 

[Electroenthusiast]

....Instead of just saying 'you're wrong' you should post a link to the appropriate site that shows how the notation should be.

I'll wager that most digital pundits weren't even aware that there was an SI prefix system adopted.
I'd call that a major marketing fail for SI.

I checked this with a file on my system... Even the microsoft windows isn't following this.
Size: 3.98 MB (4,175,998 bytes)
Size on Disk: 3.98 MB (4,177,921 bytes)
But, i got another ques... why this diffs 'size and size on Disk??

....

You has 1 start bit, 7 or 8 data bits, 0 or 1 parity bit and 1, 1.5 or 2 stop bits. Two stop bits were necessary so the motor had time to stop.

The usual 1 start, 8 data and 2 stop so baud/10 sometimes is characters per second.

The start/stop bits are Overhead.

There are lots of overheads in networking protocols (IP and TCP headers from your machine, repackaged with PPP and ATM headers from the DSL router to the ISP etc) which mean that the entire bandwidth is not exclusively used for "downloading data". Nigel refers to these as "supporting bits".

Video:
https://www.youtube.com/v/Ve7_4ot-Dzs

Considering data transfer using internet, assume that i'm downloading 1MB (1000kilo bytes) file, whats the ratio of supporting bit's versus File Size. What i'm asking is, will every bit that is to be transferred have a set of supporting bits?/ will the complete 1MB file have a single set of supporting bits?.

 

[edeca]

Considering data transfer using internet, assume that i'm downloading 1MB (1000kilo bytes) file, whats the ratio of supporting bit's versus File Size. What i'm asking is, will every bit that is to be transferred have a set of supporting bits?/ will the complete 1MB file have a single set of supporting bits?.

Disclaimer: there is no such thing as a perfect world, the numbers below vary based on lots of things. I've picked some common values.

On your local network it is quite common for 1500 data bytes to be sent in one Ethernet frame. If we pick the very common TCP/IP to send our data then at least 40 bytes will be headers (the "overhead"). These headers tell the network things like:

- where the packet is going (address and port)
- where it came from (address and port)
- which other packets it relates to
(etc)

That leaves 1460 bytes for your application. UDP encapsulation has a useful diagram (for UDP), see how the lime green data has extra data added at each step.

Depending on the application there will be further overhead, such as HTTP headers, FTP commands, Samba messages etc. If we ignore those for now, let's assume we are going to send 1048576 bytes across the network.

If we are able to use the maximum 1460 bytes available in each packet, that leaves us:

1048576 data bytes total / 1460 data bets per packet = 718.2 packets

From this, we know that we will have a certain amount of overhead:

718.2 packets * 40 bytes per packet overhead = 28728 bytes overhead

So to send your data we needed to break it into ~718 "chunks" to put it into packets and also send 28728 bytes extra to make sure it all arrives OK.

And if we know that each packet is 40 bytes overhead and 1460 bytes data, we can see that there is approximately 2.6% overhead in total for your data. Obviously this is the lowest possible estimate, as there will always be more in the "real" world.

Note that this only covers your local network and it is already quite complicated. To send data outside your network and to the internet this is all repackaged and sent off in different formats ("encapsulations"). Each time this happens there is more overhead and therefore less space available for your data. Hopefully this post gives you enough to think about what overheads there are.

No post would be complete without a reference to the OSI model, see https://en.wikipedia.org/wiki/OSI_model for a thoroughly confusing Wikipedia entry!

The most important thing to remember here is that the overhead in sending data in the networking sense is not impacted by what electrical bits are transmitted across the wire, wifi, infrared port etc. This is all very important but happens at a lower level that doesn't implicitly affect the kind of "throughput" your post is talking about.