People Against FCoE, Its Loud and Proud — Part 1

For those of you who are think­ing of rush­ing into FCoE and the Cisco Nexus switches you might want to think again. There are a few people com­ing out against FCoE and ask­ing why is any­one both­er­ing ? Should I be put­ting it in my design port­fo­lio ? Is FCoE a done deal just because Cisco is throw­ing it weight around, or does it have merit.

In fact after some research , there are so many neg­at­ives I had to put them in mul­tiple parts.

Dell is in favour of iSCSI and thinks that FCoE is ” about as stu­pid as a tech­no­logy could be. Vendors mak­ing it just want to gen­er­ate higher mar­gins with a new technology.”

A very old post­ing (from Aug 29 2007!) point­ing out the FCoE was waste of space. Its an eru­dite and robust art­icle cov­er­ing both sides of the debate, with FCoE def­in­itely los­ing. Catfight Over FCoE.

Reasons not to use FCoE include:

* FCP end­points are inher­ently cost­lier than simple NICs – the cost argu­ment (ini­ti­at­ors are more expens­ive)
* The credit mech­an­isms is highly unstable for lar­ger net­works (check switch vendors plan­ning docs for the net­work dia­meter limits) – the scal­ing argu­ment
* The assump­tion of low losses due to errors might rad­ic­ally change when mov­ing from 1 to 10 Gb/​s – the scal­ing argu­ment
* Ethernet has no credit mech­an­ism and any mech­an­ism with a sim­ilar effect increases the end point cost.
* Building a trans­port layer in the pro­tocol stack has always been the pre­ferred choice of the net­work­ing com­munity – the com­munity argu­ment
* The “per­form­ance pen­alty” of a com­plete pro­tocol stack has always been over­stated (and over­rated). Advances in pro­tocol stack imple­ment­a­tion and finer tun­ing of the con­ges­tion con­trol mech­an­isms make con­ven­tional TCP/​IP per­form­ing well even at 10 Gb/​s and over.
* Moreover the mul­ticore pro­cessors that become dom­in­ant on the com­put­ing scene have enough com­pute cycles avail­able to make any “off­load­ing” pos­sible as a mere code restruc­tur­ing exer­cise (see the stack reports from Intel, IBM etc.)
* Building on a com­plete stack makes avail­able a wealth of oper­a­tional and man­age­ment mech­an­isms built over the years by the net­work­ing com­munity (rout­ing, pro­vi­sion­ing, secur­ity, ser­vice loc­a­tion etc.) – the com­munity argu­ment
* Higher level stor­age access over an IP net­work is widely avail­able and hav­ing both block and file served over the same con­nec­tion with the same sup­port and man­age­ment struc­ture is com­pel­ling – the com­munity argu­ment
* Highly effi­cient net­works are easy to build over IP with optimal (shortest path) rout­ing while Layer 2 net­works use bridging and are lim­ited by the logical tree struc­ture that bridges must fol­low. The effort to com­bine routers and bridges (rbridges) is prom­ising to change that but it will take some time to finalize(and we don’t know exactly how it will oper­ate). Untill then the scale of Layer 2 net­work is going to ser­i­ously lim­ited – the scal­ing argument

Note also that Silvano Gai (of Nouva now acquired by Cisco and chief cheer­leader for FCoE) try­ing to spike the pro-​​iSCSI debate by push­ing some pro­cess com­plaint into the argu­ment about stand­ards. Bad form there.

My favor­ite bit

The only argu­ment I have heard for con­tinu­ing any devel­op­ment in the FC pro­to­cols at all, whether that devel­op­ment is done at T-​​11 or at IETF, is to provide the means to wean all the crack addicts in the Global 2000 off of FC fab­rics alto­gether, and as soon as humanly pos­sible. FC SAN is simply the most expens­ive way to host data that was ever inven­ted. No sur­prise that it came to mar­ket at a time when every­one was sus­pend­ing dis­be­lief and invest­ing in dotcoms.

I agree com­pletely with this point.

fcoe​.com

I had a look at a Web site fcoe​.com, with lots of nice inform­a­tion. However, it too is owned by Nouva Systems and thus Cisco.

Is there any­one else but Cisco sup­port­ing FCOE ?

Note: Apologies for the ads in that screen­shot but the Network Solutions whois server is abso­lutely criminal.

Please rate this post:

(No Ratings Yet)

 Loading ...