Prev: Belkin Wireless Router Connection Problems/Hanging/Freezing
Next: Top Ten Reasons For Going Mobile in 2010
From: Jeff Liebermann on 26 Dec 2009 17:24
On Sat, 26 Dec 2009 19:45:10 +0000, alexd <troffasky(a)hotmail.com>
wrote:

>On second thoughts, two different types of iperf floating around the
>internet could lead to much confusion when trying to compare speed test
>results.

Well, since he's apparently benchmarking some device, it must mean he
actually has something working. There's hope, methinks, maybe.

Perhaps it would be more appropriate to ask on the iPerf mailing list:
<https://lists.sourceforge.net/lists/listinfo/iperf-users>
Oh, he already posted a question and got a mostly wrong answer. Oh
well.

Pretend I didn't mention IEC 60027-2 A.2 which uses kibi, mebi, and
gibi bytes.
<http://members.optus.net/alexey/prefBin.xhtml>
<http://physics.nist.gov/cuu/Units/binary.html>

However, he wanted to know why iPerf did it both ways. The first
version of iPerf was scribbled in May 2001. The ISO released the
binary prefix standards in 1998, which are generally ignored by the
industry to this day. Until the failure to use kibi, mebi, and gibi
bytes is made an international crime punishable by being forced to
read the entire standard from cover to cover, the choice of prefixes
are those of the author.

--
Jeff Liebermann jeffl(a)cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558
From: Jeff Liebermann on 26 Dec 2009 17:42
On Sat, 26 Dec 2009 02:10:07 -0800 (PST), karthikbalaguru
<karthikbalaguru79(a)gmail.com> wrote:

>I wonder why Iperf uses 1024*1024 for megabytes
>and 1000*1000 for megabits ?

For some entertainment value, try TTCP:
<http://www.pcausa.com/Utilities/pcattcp.htm>
and see how the results compare.

--
Jeff Liebermann jeffl(a)cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558
From: Mark Moulding on 26 Dec 2009 21:21

"Jeff Liebermann" <jeffl(a)cruzio.com> wrote in message
news:vqjcj55sc1ee5h2i29d8cbfe6vg7is35mi(a)4ax.com...

> For example: 100'000'000 bytes is not equal to 100 Mbytes but
> to 100'000'000/1024/1024 = 95.37 Mbytes.

Really? I always thought that (in digital-land) since 1 Kbyte = 1024 bytes,
a Mbyte is 1024*1024 = 1,048,576 bytes. And a Gbyte = 1024 (1k) mBytes =
1024^3 = 1,073,741,824 bytes. In other words, "digital" SI prefixes are a
little bit larger than "engineering" ones, and conveniently follow
power-of-two groupings (kilo=2^10, mega=2^20, giga=2^30, tera=2^40).
Wikipedia seems to agree (http://en.wikipedia.org/wiki/Binary_prefix), and
further informs me that the NIST has created a new set of prefixes that
follow this new digital meaning: "kibi" (for "kilobinary"), "mebi", etc.
--
Mark Moulding


From: Bob on 27 Dec 2009 06:24
On 26 Dec, 10:10, karthikbalaguru <karthikbalagur...(a)gmail.com> wrote:
> I wonder why Iperf uses 1024*1024 for megabytes
> and 1000*1000 for megabits ?

The real reason for this difference, which for some reason is rarely
given, is because of address decoding. If you have a RAM chip, the
memory cells are addressed by a binary address, so this leads
naturally to blocks of memory that are powers of 2. One could design
RAM chips with 1000 byte blocks say, but it makes the address decoding
unnecessarily complex. Binary addressing is the most efficient method.

For disk drives the magnetic recording medium is linear, so the block
size and cylinder/head/sector addressing can be arbitrary and not
based on powers of 2. For convenience, the sector size is chosen to be
a power of 2 to match how memory is arranged. Whether the overall
storage size of a disk is quoted in SI or K is down to preference, but
the manufacturers prefer the standard SI units.

For communications, the bit rate is determined by a clock, which is
also somewhat arbitrary and doesn't need to be based on powers of 2.
Therefore it is natural to use the standard SI units.

So in general, SI units are the preferred units. For directly
addressed RAM chips (or ROM, Flash etc) a binary unit reflects the
underlying layout, and gives an integral value. For storage media like
disks, it's a gray area, and usage depends on choice. For
communications and bit rates, SI units are normally used. In all
cases, I recommend to use the IEEE binary prefixes (KiB, MiB etc)
where appropriate to make it clear which unit is being used.
From: Rich Webb on 27 Dec 2009 06:55
On Sun, 27 Dec 2009 03:24:29 -0800 (PST), Bob
<bobcousins42(a)googlemail.com> wrote:

>On 26 Dec, 10:10, karthikbalaguru <karthikbalagur...(a)gmail.com> wrote:
>> I wonder why Iperf uses 1024*1024 for megabytes
>> and 1000*1000 for megabits ?
>
[snippety snip]
>So in general, SI units are the preferred units. For directly
>addressed RAM chips (or ROM, Flash etc) a binary unit reflects the
>underlying layout, and gives an integral value. For storage media like
>disks, it's a gray area, and usage depends on choice.

The more cynical among us may contend that usage depends on marketing.

Since, for a given quantity, an SI enumerated size is "bigger" than the
equivalent binary size, which sells more? A 500 SI-gigabyte drive or a
466 binary-gigabyte drive? Or even a 480 binary-gigabyte? Who looks at
the fine print?

Some of us <cough> are old enough to remember when 64 Kbyte machines
were the top of the line. Naturally, some advertising copy referred to
*their* machines as "65 Kbytes!" and a few even noted that 65,536
conventionally rounds to "66 Kbytes" and, yes, advertised them that way.

--
Rich Webb Norfolk, VA
First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4
Prev: Belkin Wireless Router Connection Problems/Hanging/Freezing
Next: Top Ten Reasons For Going Mobile in 2010