bits and bytes

[helltek]

Hi guys n gals. I read somewhere in these posts about converting bits per second to bytes per second network throughput as dividing by 8. Yes, there are 8 bits in a byte, but 8,000bps does not transfer 1,000 bytes in one second. There is overhead that needs to be considered. Just like in RS-232 serial is start and stop bit (10 bits per byte), there is also overhead in Ethernet. I'm not an expert on this subject, but I think that the division may be closer to divide by 10 instead of divide by 8.

 

[ex_bubblehead]

Just as a rule of thumb use a divisor of 8, then reduce by 15-20% (roughly) for protocol overhead. Gets you close enough.

 

[helltek]

Thanks for your correction. Still it comes to more or less the same final result, so when the specs show bits per second, just drop the last zero. 100kbps = 10kB/s. That's when there are no other delays.

This may be the least interesting thread on this site, but programmers today repeat problems that were resolved 25-30 years ago. Of course these programmers were not born 30 years ago, and they don't read and learn. "If you don't learn history, you will repeat it".

 

[bill001g]

There is no way to predict the overhead especially you use the term the way the common users do. You example of a serial cable is to low a level to look at this. The general user is talking about how long does it take to copy files. Sure you could get some generic number for say a cable modem and a dsl connection or a OC12 packet over sonnet connection but they are not even the same.

When you look at the real world the users are considering all the IP headers overhead also. All they care about is how long does it take to get the actual data. So if we look at a very simple case where we look at a simple ipv4 header which is generally 20 bytes. If your packets are sent in 200 byte chunks you have 10% overhead but if you send them in 1000 byte chunks the overhead is 2%. So there is no way to get a number even in the simple case. Now think of all the other possible things. Maybe you have encryption headers or you are using IPv6 instead or you have vlan tags on the packets.

It is way to simplistic to think that you can divide by 10 to get the transmission rates in bytes....talking about old data that is why this myth of divide by 10 still exists it is based on old serial transmission protocols.

 

[helltek]

You're correct. There is whole lot more than theoretical media speed and file size. I started this thread in the wrong place. I should have find the thread where someone compared USB2, USB3 to Ethernet NAS. They used division by 8. I just wanted to point out that they need to consider overhead. Dividing bps by 8 gives incorrect expectation that can not be achieved.