Max throughput for 2gb and 4gb SAN Switch ports?

[gurdeep_2]

What are the theoretically max throughputs for a 2gb and 4gb SAN switch ports?

The SAN switch is a Brocade 4gb/32-port switch.

This switch is running my enterprise backup infrastructure: Windows 2003, Neoscale Encryption, HP-EML Tape Library.

[Jeff_Traigle · Accepted solution]

Pretty simple calculation. Divide by 10. (Each byte is 10 bits after encoding.)

2Gbps = 200MBps
4Gbps = 400MBps

[krusty]

Since FC is duplex, I would contend double what Jeff had, so:

2Gb = 400 MB/sec
4Gb = 800 MB/sec

Cheers,

Curt

[Alzhy]

2Gbit == 160 MiB/sec
4Gbit == 320 MiB/sec

where MiB is 1024^3 bytes (aka Megabytes)

[Johan Guldmyr]

But what about the 8b/10b encoding? Wouldn't that make it drop 20% of theoretical max as well?

I've heard the same as what Jeff has written.
2gigabit/s -> 200megabyte/s
etc

[Alzhy]

Oh c'mon:

1 byte = 8 bits.

Do the Math. There is also the overhead.

Unless you folks are saying that:

MByes or Megabytes per second = 1,000,000 bytes/sec?

[Uwe Zessin]

There is a little more than 'simple' maths.

On 1GBit Fibre Channel, the bit clocking rate is 1.0625 GigaBit per second.
Each 8-bit data byte from the computer is converted into a 10-bit 'transmission character'. The reason is that on a serial channel, you must not send too many 1 or 0 bits in a row. The additional 2 bit allows the sender to introduce bit changes early enough.

This also helps the receiver to keep in synch with the transmitter as here is no separate clock signal. There are additional rules, but I try to keep it easy.

So 100 data MegaBytes per second become 800 data MegaBits per second which become 1,000 transmission bits per second.

What about the additional 0.0625 GigaBit you ask? Well that is for the management overhead like header, addresses, checksums.

It is true that these values are based on 1000, not 1024, but that is how it is announced everywhere and, no offense, I don't think it makes sense to start calculating with 1024.

[Alzhy]

I hear ya Oh Great Uwe.

So it follows the normal industry storage sizing of a MB meaning 1,000,000 bytes. So a GB is 1,000 MB and 1 TB is 1,000 GB.

And the poor end user who slaps a 2TB HDD on his PC finds out as Micro$oft reports - the 2TB HDD Capacity being reported as 1.8TB!

;^))

I do now notice most Hard Disk labels now explitcitly note a GB as 1 billion bytes to avoid "complications"

[Uwe Zessin]

Oh, thanks. I'm not sure if that was cynical or meant as a compliment ;-)


(Part of) the problem is that the industry, after all these years, has not managed to educate itself and its customers about the difference between what I call "HardWare Gigabytes / HWGB / 1000" and "SoftWare GigaBytes / SWGB / 1024" and teach their products to clearly indicate what is shown.

I don't know how often I've written or talked about the difference - sorry if I sound like a broken record.


Kids, if you don't know what I mean by 'broken record', don't bother Google - just click here ;-)

http://en.wikipedia.org/wiki/Gramophone_record
http://idioms.thefreedictionary.com/sound+like+a+broken+record

[Alzhy]

Well - in my shop, it is stipulated in all procedures, contracts, vendor communications, etc -- that all storage specifications are to be based on the fact that 1MB means 1024x124 bytes.

All the tools and storage suites we use and even OS utilities dealing with Storage Follow this norm anyway -- so why follow the storage vendor "disk" capacity standards?

[Jeff_Traigle]

My simple calculation was per SNIA definition. The question was about throughput, which is "the measurement of the achievable or usable data traffic in one direction only." Therefore, duplex is irrelevant in determining throughput. Distance, media, latency, and flow control can all affect throughput, but the theoretical maximum throughput (as requested) assumes no impact of any of these factors.

Bandwidth would be twice throughput for full duplex, which is what Curt alluded to.