[c-nsp] When will SFP+ 10GBase-T optics be available?

[Phil Bedard]

A couple years ago we built a new datacenter and we opted to not put
structured cabling in everywhere and backhaul things to EoR like we did
previously because of the uncertainy of where host 10GbE was going.   At
the time no one really had a 10GBaseT switch out except for Arista and we
were looking at using FCoE (a whole nother topic) and the couple vendors
supporting FCoE had no intentions at the time of supporting it on their
upcoming 10GBaseT switches (Cisco 2232TM FEX).  Nowadays Intel is pushing
using a software FCoE initiator (Open FCoE) with NIC offload just like
some iSCSI solutions, and they are supporting it on their LOM 10GBaseT
NICs.   I believe VMWare ESX has a software FCoE initiator now but don't
quote me on that. 

Also when we looked at the number of potential 10G ports the power usage
at the time per 10GBaseT port was 10x that of a fiber/twinax solution, but
the power numbers have come down quite a bit and on the switch side there
isn't really much difference these days between SFP+ and 10GBaseT.

As for 10GBaseT SFP+, I've also been told it will never happen, so you'll
be forced to choose between using fiber/twinax or cat6 at least for the
upcoming future.  

Phil  

On 4/21/12 3:17 AM, "Eric Rosenberry" <eric.rosenberry at iovation.com> wrote:

>inline
>
>On Fri, Apr 20, 2012 at 1:08 AM, Phil Mayers
><p.mayers at imperial.ac.uk>wrote:
>
>> On 04/19/2012 11:57 PM, Eric Rosenberry wrote:
>>
>>> I have hosts and storage arrays arriving that are coming with 10GBase-T
>>> ports onboard (and no SFP+ ports).  This makes it very hard to hook to
>>>my
>>> SFP+ *only* switches.  ;-)
>>>
>>
>> Really? Hmm. Interesting - what equipment is this?
>>
>
>Dell MD 3620i storage arrays.  Don't ask me why they made the decision to
>be *only* 10GBase-T.  I suspect it may have something to do with their
>target audience not being comfortable with deploying fiber?
>
>Newly launched Dell 720 series mainline servers with the Intel 10 gig NIC.
>
>Newly launched Supermicro platforms with 10GBase-T Intel controllers
>onboard.
>
>Note that as others I think eluded to on the list, it is the Intel X540
>controller that both of these two sever platforms are using:
>http://newsroom.intel.com/community/intel_newsroom/blog/2012/03/06/chip-sh
>ot-intel-ethernet-controller-x540-brings-10gbe-to-the-masses
>
>I think this controller has kicked off the first serious wave of servers
>coming with 10GBase-T.
>
>
>> There does not seem to be a (genuine, unbiased) consensus about whether
>> 10GbaseT or SFP+ is going to "win out" for host connectivity. We have
>> dithered on this, but have moved in the direction of SFP+ on the basis
>>that
>> server manufacturers seem to be shipping SFP+ 10G kit, and the cheap 1U
>>ToR
>> switches are largely SFP+
>
>
>My personal opinion is that 10 gig to the majority of servers will only
>become popular once it is in a 10GBase-T format.  Fiber is a pain, and
>expensive (especially including the optics).  Twinax is cost effective,
>but
>fragile and compatibility has sometimes been an issue.  Not to mention
>that
>it sucks that it only comes in a handful of pre-made lengths (CAT6a is
>much
>more widely available).  I see Twinax as a bridging technology until
>10GBase-T is cost effective, low power, and bulletproof in implementation.
>
>-Eric
>
>-- 
>*Eric Rosenberry*
>Sr. Infrastructure Architect // Chief Bit Plumber
>_______________________________________________
>cisco-nsp mailing list  cisco-nsp at puck.nether.net
>https://puck.nether.net/mailman/listinfo/cisco-nsp
>archive at http://puck.nether.net/pipermail/cisco-nsp/