Breaking News
Comments
Newest First | Oldest First | Threaded View
Page 1 / 3   >   >>
gbaggett750
User Rank
Rookie
re: Slideshow: Intel Lights 100G Interconnects at Facebook Summit
gbaggett750   1/30/2013 6:50:03 PM
NO RATINGS
I'm sorry but this announcement isn't as exciting as it should be...maybe it's because I am not in a good mood today. 100G in a single optical module. This is all fine and good if you desire point-to-point communication only. Separating CPU, RAM and storage by using fiber links only invites disaster as you now have more componets that can fail and therefore MTTF goes down. More importantly is when are the major router/switch box producers going to get involved? Did anyone talk with them? Without them, as far as I am concerned, this is pointless. Can you see Cisco, Juniper or Big Iron putting a PCIx16 card slot into their machines so that 100G switching/routing can occur? or even 4x25G? This is where the explosion in the computing centers will occur. Without I/O to the world, the compute centers become data bound and do not need to expand their processing capacity. I cannot see a switch using 4x25G ports unless multiple switches are run in parallel or one big switch with 4 fabrics. Four times the cost for implementing switching or wait for a re-designed switch. I'll wait. On the other hand, I envision a complete redesign of switch fabric to take advantage of this 100G I/O. Can you say photonic fabric? Now this would be an important and meaningfull announcement.

Chris Cole
User Rank
Rookie
re: Slideshow: Intel Lights 100G Interconnects at Facebook Summit
Chris Cole   1/26/2013 6:57:25 PM
NO RATINGS
Hello Shmuel, The answer to your question about the Intel Si Photonics module in the Open Compute Summit demo, based on the type of rack interconnect shown, is that it has four parallel 25G modulators driving four parallel single-mode fibers. This is the same approach as proposed for standardization in the IEEE 802.3bm 40 Gb/s and 100 Gb/s Fiber Optic Task Force by three companies (Avago, Luxtera, Oclaro), for up to 500m reach data center application, and referred to as PSM4 (parallel single-mode 4-channel). Each company showed the technical feasibility of this approach based on their internal and unique technology, both Si Mod and InP DML based. What they have failed to do so far is demonstrate that there is an economic advantage and broad market potential. Many end users, including carriers and some IDC operators prefer duplex SMF for longer reach applications. This is enabled by WDM, for example the existing IEEE LR4 standard, the 2010 Intel announcement, recent IBM IEEE proposal, the Kotura announcement (see link provided by Rick) which they proposed to the IEEE, and many others. It is therefore not clear why on this thread commenting about the Intel demo at the Open Compute Initiative, there is a detailed description of how to set the frequency of each color {wavelength}. Perhaps a technical conference paper about the earlier work accidently got into the Summit marketing package.

resistion
User Rank
Manager
re: Slideshow: Intel Lights 100G Interconnects at Facebook Summit
resistion   1/26/2013 3:49:22 AM
NO RATINGS
@Justin, just saw your post. Glad we were in sync.:)

resistion
User Rank
Manager
re: Slideshow: Intel Lights 100G Interconnects at Facebook Summit
resistion   1/26/2013 3:47:23 AM
NO RATINGS
A silicon Raman laser still requires optical pumping, cannot be activated by on-chip electronics. So they're bonding III-V.

rick merritt
User Rank
Blogger
re: Slideshow: Intel Lights 100G Interconnects at Facebook Summit
rick merritt   1/25/2013 9:48:30 PM
NO RATINGS
For a broader analysis of this area including news of a pending foundry deal for Kotura see: http://www.eetimes.com/electronics-news/4405636/Kotura-preps-fab-deal-amid-silicon-photonics-surge

rick merritt
User Rank
Blogger
re: Slideshow: Intel Lights 100G Interconnects at Facebook Summit
rick merritt   1/25/2013 9:48:16 PM
NO RATINGS
For a broader analysis of this area including news of a pending foundry deal for Kotura see: http://www.eetimes.com/electronics-news/4405636/Kotura-preps-fab-deal-amid-silicon-photonics-surge

shmuel levy
User Rank
Rookie
re: Slideshow: Intel Lights 100G Interconnects at Facebook Summit
shmuel levy   1/24/2013 7:39:31 AM
NO RATINGS
Hi Chris, is the Intel modulator has 25G per channel?

shmuel levy
User Rank
Rookie
re: Slideshow: Intel Lights 100G Interconnects at Facebook Summit
shmuel levy   1/23/2013 8:24:41 PM
NO RATINGS
So the Intel device has 4 fibers each at 25G?

Chris Cole
User Rank
Rookie
re: Slideshow: Intel Lights 100G Interconnects at Facebook Summit
Chris Cole   1/23/2013 1:09:31 PM
NO RATINGS
Another example of the announcement being light on content and heavy on marketing is the liberal use of adjectives to specify performance, instead of the more traditional numbers. For example, the modules are claimed to operate at “very low power levels”. At least for this spec we can quantify the adjectives because the modules were visible on the cards and had hefty heat sinks mounted on them. The heat sink size puts the module power dissipation above 3W, possibly 4W. This is similar to what other 100G board mounted optics dissipate today. (Also, earlier comment stated 4x15.2G for the 2010 announcement; it should have stated 4x12.5G).

Chris Cole
User Rank
Rookie
re: Slideshow: Intel Lights 100G Interconnects at Facebook Summit
Chris Cole   1/23/2013 12:33:38 PM
NO RATINGS
The 2010 Intel announcement was a 4x15.2G WDM Muxed/Demux chip set that carried 50Gb/s per fiber. The announcement last week is no longer WDM, and only carries 25Gb/s per fiber, so that’s a drop in fiber capacity by a factor of 2x. This research has been going on for the past decade and the track record so far of commercializing technology is not great. If we extrapolate the trend, in 2014 there will be a big media event announcing a great new Si Photonics chip set which will carry 10Gb/s per fiber. The announced 100G server to TOR switch application, based on Intel server roadmap, will not be need until 2017 and appreciable 100G server I/O volume won’t happen until 2020. So the good news is that gives 4 to 7 more years to iron out the bugs in the announced chip set. Unfortunately by then, 4x25G will not be the architecture of choice for I/O as the industry moves to 50Gb/s per lane and fiber. Perhaps better coordination with the guys that design products that pay the bills at Intel might help focus the photonics research towards something that would help the company.

Page 1 / 3   >   >>


Flash Poll
EE Life
Frankenstein's Fix, Teardowns, Sideshows, Design Contests, Reader Content & More
Rishabh N. Mahajani, High School Senior and Future Engineer

Future Engineers: Don’t 'Trip Up' on Your College Road Trip
Rishabh N. Mahajani, High School Senior and Future Engineer
1 Comment
A future engineer shares his impressions of a recent tour of top schools and offers advice on making the most of the time-honored tradition of the college road trip.

Max Maxfield

Juggling a Cornucopia of Projects
Max Maxfield
6 comments
I feel like I'm juggling a lot of hobby projects at the moment. The problem is that I can't juggle. Actually, that's not strictly true -- I can juggle ten fine china dinner plates, but ...

Larry Desjardin

Engineers Should Study Finance: 5 Reasons Why
Larry Desjardin
34 comments
I'm a big proponent of engineers learning financial basics. Why? Because engineers are making decisions all the time, in multiple ways. Having a good financial understanding guides these ...

Karen Field

July Cartoon Caption Contest: Let's Talk Some Trash
Karen Field
128 comments
Steve Jobs allegedly got his start by dumpster diving with the Computer Club at Homestead High in the early 1970s.

latest comment mhrackin Where's the "empty bin" link?
Top Comments of the Week
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed

Datasheets.com Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)