Where servers, storage and networking combine to form Voltron.
February 9, 2015
Filed under Uncategorized
I have to admit to being a little thrown off by the numbers. It starts off well enough with the premise that Fibre channel provides about 100MBps of throughput per gig. The author then proceeds to show how Ethernet provides 125MBps (true 1Gbps) by upping the baud rate from 1Gbaud to 1.25GBaud. But wait, 1G FC doesn’t use 1GBaud, it uses 1.0625GBaud. Following the same math used to show Ethernet throughput, 1G FC should provide 106.25MBps, not 100MBps.
OK, so we’re only 6.25MBps off here. The general statement of 1G FC provides ~100MBps throughput is still close enough to call it a wash. However, this discrepancy is exaggerated for higher speeds of FC where the author never actually runs the numbers. At the end, the author makes a pretty big point about how 128G FC is only 300MBps faster than 100G Ethernet. Knowing that the numbers for 1G FC are off, lets look at the actual numbers for 128G FC.
Throughput = (Data Bits / Encoded Bits) * Baud. (64/66)*112.2G=108.8Gbps. Divide by 8 for bits->bytes, multiply by 1000 for GBps->MBps, 108.8Gbps/8*1000=13600MBps. 13600 is a fair way off from 12800. This makes the difference between 128G FC and 100G Ethernet 1100MBps, not 300MBps.
I think the basic idea that speed in FC and Ethernet are measured differently is absolutely worth knowing. It’s also probably a good idea to remember that 1G of FC provides ABOUT 100MBps of throughput. But if you’re going to report on the numbers in a side-by-side comparison, it is important to consider the “ABOUT” in that statement. Or am I missing some sort of overhead in FC transmission that isn’t reported here?
On the Fibre Channel side, I’m not sure what the cause is for the small baud discrepancy. Perhaps another signalling/protocol overhead that they bump up slightly to make room for FC frames (looks like about 6.25%, and it’s consistent).
For transfer speed (measured in megabytes per second) I use the transfer speeds what the vendors and the T11 standard (http://www.t11.org/ftp/t11/pub/fc/pi-5/11-011v0.pdf page 30 of the PDF) all report. They all says 100/200/400/800/1600 etc. for transfer speeds, so I’m going with those transfer speeds.
Just purusing some docs, I can’t find the source of the discrepancy. Great question, I’ll look into it more.
Fascinating, I hadn’t realized that those numbers were from the standard itself. Obviously I would hope that the standards body wouldn’t simply “forget” 6.25%, so I’m sure there is a missing variable there. Also, my apologies for continually referencing “the author” without bothering to notice that you are the author of the video. Great stuff, keep up the good work!
Fill in your details below or click an icon to log in:
You are commenting using your WordPress.com account. ( Log Out / Change )
You are commenting using your Twitter account. ( Log Out / Change )
You are commenting using your Facebook account. ( Log Out / Change )
You are commenting using your Google+ account. ( Log Out / Change )
Connecting to %s
Notify me of new comments via email.
Blog at WordPress.com.