neodata686 All American 11577 Posts user info edit post |
According to wikipedia throughput is the:
Quote : | "In communication networks, such as Ethernet or packet radio, throughput is the average rate of successful message delivery over a communication channel." |
So for example with the wireless g protocol you get a 54Mbit/s max net bitrate, but only a 19 Mbit/s throughput. So does this mean on average you should get 19Mbit/s with wireless g?
Also wireless n is said to have a 600Mbit/s max net bitrate, but only a 74Mbit/s throughput bitrate. Why the large difference? Is max net bitrate the technical max bitrate without packet losses? and throughput is the tested average with losses? Just wondering. Maybe somehow who knows about this can clear things up for me.
And when you're speaking in terms of data transfer rate, that would mean the throughput with losses from the physical layer, overhead, etc...Right?12/14/2008 10:37:09 PM |
moron All American 34142 Posts user info edit post |
There's a different between MB (capital "B") and Mb (lowercase "b"). Make sure youre not mixing it up. One is Bytes and one is bits. 12/14/2008 10:56:32 PM |
amac884 All American 25609 Posts user info edit post |
There's a different between MB (capital "B") and Mb (lowercase "b"). Make sure youre not mixing it up. One is Bytes and one is bits.
[Edited on December 14, 2008 at 11:02 PM. Reason : ^ ah, he beat me] 12/14/2008 11:02:04 PM |
neodata686 All American 11577 Posts user info edit post |
^,^^ i hope you're both being sarcastic.
I'm talking about data transfer, so my usage of the lowercase "b" is correct as in bits. If i were talking about data size i would use the uppercase "B" for bytes.
I even typed "Mbit" instead of "Mb." 12/14/2008 11:04:11 PM |
moron All American 34142 Posts user info edit post |
^ the conversion from bytes to bits is approximately multiplying by 8, which is what 600/8=~74 and 54/8=~8 but maybe you were talking about those TURBO g spec which uses 2 G channels to double the speed...?
I just figured you misreading your sources or something. 12/14/2008 11:08:47 PM |
neodata686 All American 11577 Posts user info edit post |
lol no. i think i learned the difference between bits and bytes when my dad downloaded my first apogee game on our 28k modem in the early 90's.
Data transfer rate is in bits, and data size is in bytes. I was asking about throughput, not how many bits in a byte.
http://en.wikipedia.org/wiki/54g
I was asking about the difference between throughput and max bitrate, and how they related to packet loss, overhead, etc over a wireless network. I'm just wondering how they got that 19Mbit/s throughput for wireless G, compared to a max of 54, and why in the world wireless N has a max of 600Mbit/s, and a throughput of 74Mbit. (maybe it's a typo, because as you pointed out, 600/8 is 74). 12/14/2008 11:16:22 PM |
Aficionado Suspended 22518 Posts user info edit post |
well there is a citation needed in that chart 12/14/2008 11:19:51 PM |
neodata686 All American 11577 Posts user info edit post |
Yeah meaning the 74 is probably a typo. 12/14/2008 11:28:12 PM |
moron All American 34142 Posts user info edit post |
^^^ It just looks like they're using "throughput" to mean "average throughput" when you can measure "maximum throughput" "minimum throughput" etc. Bit rate is more of a unit of measurement (kind of like inches are a measurement of length, bit rate is a measurement of throughput), but they are often used interchangeably.
I wouldn't think though the average throughput of a wireless network would be 1/8th of its max, that seems wrong to me.
[Edited on December 14, 2008 at 11:31 PM. Reason : ] 12/14/2008 11:31:00 PM |
neodata686 All American 11577 Posts user info edit post |
^There's a goodput too! Application layer throughput. 12/14/2008 11:33:02 PM |
cdubya All American 3046 Posts user info edit post |
To answer the original poster:
Quote : | "So does this mean on average you should get 19Mbit/s with wireless g?" |
imho- there are far too many variables to make a generalization like that.
Just a few considerations that come to mind- there are tons more to dig into: buffer method/size rtt csma/cd behavior transport packet loss recovery/retransmission methods window sizes stack/transport optimizations
If we're strictly discussing tcp, have a look at rfc1323,2018, and 2581. Some material is a bit dated, but they still bring to light the challenges involved and some of the standardized solutions.
Hope this helps!
Edit: finding myself to be ignorant in the L1/2 considerations of this questions, one google query yielded: http://pdos.csail.mit.edu/papers/jbicket-ms.pdf Pretty interesting read!
[Edited on December 15, 2008 at 12:17 AM. Reason : l1/2]12/15/2008 12:09:36 AM |
neodata686 All American 11577 Posts user info edit post |
^thanks! What i'm reading over now. I just didn't think they could make a generalization and say a wireless network will always get a throughput of 19Mbit/s, when obviously this is dependent on so many things. 12/15/2008 12:15:04 AM |
cdubya All American 3046 Posts user info edit post |
^completely agreed- 19mbs is a situational stat for sure. 12/15/2008 12:18:12 AM |
mellocj All American 1872 Posts user info edit post |
consumer wireless 802.11 stuff never works up to the full capacity like wired ethernet can. If you are getting 19Mbit/s of real throughput on 802.11g, I would say that is typical or pretty good. 12/15/2008 8:45:18 AM |