7/12/2011
Hitting on the comparison between FC and SAS. It was reallydifficult in the beginning – so many concepts I know about, or heard about butjust don’t really understand. And so many protocols. FC has at least 7– it issupposed to have only five, but I don’t know where the FC-AL protocol come from,well, should be a sub-protocol of FC-2. And so is the FC-SW protocol.
Also I need to know SAS. Its protocol looks not many – allthree sub-protocols are incorporated into the SAS standard. But I don’t knowwhy after SAS 2.1, T10 divided the protocol into two documents: SAS, and SASProtocol Layer. Both’s names are greatly ambiguous – could it be that T10 folksthought a single SAS protocol looks too simple hence tried to make it lookcomplicated?
Anyway, comparing studying SCSI protocol, FC and SAS aremuch easier – they are just carrier protocols. FC puts a lot controllingmechanism into its frame header by which it can play a lots of tricks whichbring it the reputation of high reliability and high performance but also makeit unfamiliar to most IT ppl. SAS is a even simpler protocol. Like FC, It isalso a carrier layer dealing with the lower 4 layers of OSI model.
7/17/2011
Done the paper of FC vs SAS. Really tedious work – manythings I actually don’t have hands on experiences but have to dig in. althoughmy experience with hw, sw and sys seems not too low to understand the intentsthat the standard tries to express, but it is indeed a tiring work, and hard toremember for those unfamiliar stuff – even I kinda understood them at the timewriting the paper.
No sweet w/o sweat. Vice versa, if you spill your sweat in aproper way, you will be assured to get some sweet. So was my paper writing. Afterfinishing this battle, I found I gained much more sense about the protocols. Previouslyit is just hard for me to grab the essence of some descriptions in a groups ofassociated protocols. But now, after comparing quite a number of SAS and FCprotocols, as well as reading a bunch of articles, now I’ve got some feelingabout these bibles in technical territory. I guess a clergy could have thesimilar feeling when he starts to understand the bible bit by bit, day by day…
Now get to the tech stuff. The comparison shows FC is reallyat some inadvantageous position compared with SAS. The only aspect it beats SASis the overhead. I don’t know why SAS still sets 1024 as the limit for payload data. Whatcould be the problem if setting it to 2048? Below is from my paper:
“Since theFC_DATA Information Unit at FC-4 layer carrying the data requested by SCSI commandis directly mapped into the data field of FC-2 frame, the overhead of FC isjust the overhead of FC-2, which is calculated as below,
1 -2112/2148=1.78%
For SAS SSP,since the transport frame is encapsulated into the link layer, the calculationshould add up the overhead in both transport layer and link layer. Belowequation shows the calculation result.