Friday, March 4, 2011

10X10-2km Specification Complete and Available

The 10X10 MSA is proud to announce that the 2 km specification is complete and can be found on the 10X10 MSA website at (http://10x10msa.org/documents.htm ).  The standard was unanimously approved by the members and several companies plan to make products compliant to the standard.   Taking less than three months to create from start to finish, the specification was modeled after IEEE specifications and includes illustrative link budgets and information on how to implement the solution in the CFP Form Factor.
Since the 2km standard is complete, the MSA has agreed to develop more standards.  The standards that are being considered include:
1.       10X10-10km Technical Specification:  A module compliant to the 10km standard will have a larger link budget of 3dB in channel loss (10km * 0.3dB/km) vs the 2km link budget of only 0.6dB (2 km * 0.3 dB/km).  The 10 km module will interoperate with the 2km module up to a distance of 2 km.
2.       10X10-40km Technical Specification: This proposal would be for long haul applications that send signals over 40km with technology similar to 10GBASE-ER in the C-Band.
3.       10X10HD Form Factor Specification: The 10X10 MSA is investigating implementing the 10X10 specification in a smaller form factor that uses the CPPI electrical interface instead of the CAUI electrical interface.  These electrical interfaces were defined in the 100GbE standard (802.3ba) and CPPI does not require retimers, so less power is required in the module than when the retimed CAUI is used.  The power consumption goal of the new module is to consume only 6 Watts of power compared to 14 Watts for the CAUI-based 10X10 modules.  Modules that are CPPI-based will interoperate with CAUI-based modules.
4.       10X10-1310nm Specification:  The group is investigating a 2km solution based on 1310nm laser arrays that could consume less power than the 10X10-2km modules. 
So many standards, so little time.
To hear more about the 10X10 MSA and how the technology is going to be used, visit multiple presentations at OFC between March 7 and 10th at the Los Angeles Convention Center.  Meetings or talks that address various aspects of the 10X10 standards include:
1)      10X10 Open Meeting: Tuesday, March 8th at 9-10am Pacific at the Standard Hotel 12th Floor Penthouse,  550 South Flower Street at Sixth Street Los Angeles, CA. 90071   
Agenda:
9:05-9:15 – Release of 10X10 Optical Specification, Forming new MSA for Higher Density Form Factor and 10km Link standard
9:15-9:30 – Huawei presents on 1310nm 10X10 solution
9:30 -9:35 – Oplink will present on the 10X10-40km solution
9:35-9:45 - Santur presentation on High Density 10X10 Module
9:45-9:55 – Discussion on Higher Density From Factor
9:55-10:00 – Closing remarks
2)      Arlon Martin of Kotura will speak on Monday at 1:15 at the OIDA Silicon Photonics Workshop on “Developing a Roadmap for High-Performance Optical Interconnects”: http://www.oida.org/events/silicon11
3)      Bikash Koley of Google speaks on the “Where is the Optical Component Market Going?” panel on Monday at 4:00-5:30:  http://www.osa.org/corporate_gateway/corporate_programs/executive_forum/Schedule_Panels/default.aspx
4)      Scott Kipp of Brocade speaks on Monday at 4:00 “The Limits of Switch Bandwidth” at the “Meeting the Computercom Challenge Symposium II: Emerging Interconnect Technologies”: http://www.ofcnfoec.org/home/Program/Special-Symposia.aspx
5)      Donn Lee of Facebook will present Wednesday at 3:30 on “Scaling Networks in Large Data Centers”: http://www.ofcnfoec.org/osa.ofc/media/Default/PDF/2011/DataComAbstracts11.pdf
6)      Scott Kipp of Brocade speaks on Thursday from 10-12 in the “Data Center: Traffic and Technology Drivers” panel: http://www.ofcnfoec.org/Home/Exhibit-Displays-and-Activities/Market-Watch-Panel-Sessions/Panel-IV--Data-Center--Traffic-and-Technology-Driv.aspx
7)      Vijay Vusirikala of Google will speak on Thursday at 1:30 on “Drivers and Applications of Optical Technologies for Internet Data Center Networks”: http://www.ofcnfoec.org/InvitedCategory1.aspx
It’s going to be an exciting week at OFC for 10X10.
Kind regards,
10X10 MSA

Friday, February 4, 2011

The Fear of Alphabet Soup

At the open meeting of the 10X10 MSA on January 25th, the members and guests (50 in total) discussed requirements for a next generation form factor for the 10X10 that will be somewhere between the CXP and the CFP in size.  The group discussed the goldilocks problem in this presentation [http://10x10msa.org/documents/Requirements_for_Next_Generation%2010x10%20Modules.pdf] where the CXP is too small and the CFP is too big.  The new 10X10 form factor is designed to dissipate 6 Watts of power, so the CXP would be too hot while the CFP would be too cold.  Goldilocks wants one new module that is juussttt right.
The talk of several new form factors for 100GbE did not sit well with many people in the room.  End users for 10GbE understand the problem of alphabet soup all too well.  Figure 1 shows the number of ingredients that are used to create 10GbE alphabet soup. 
Figure 1: The First Serving of Alphabet Soup


Waiter: Do you want your 10GBASE-SR in an XFP, X2 or SFP+ form factor? 
User: I prefer SFP+, the X2 gave me indigestion last year.
Waiter: Would you like that with a limiting interface or a linear interface? 
User: Limiting, my network analyst told me to cut back on linear.
Waiter: Do you want that with OM3 fiber?
User: I’m on a high fiber diet of OM4. 
Waiter: Sorry, we don’t serve that here.
The problem end users face with 10GbE alphabet soup is not funny.  They have to buy, stock and track multiple replacement parts for various 10GbE switch products depending on which generation of 10GbE switch and blade they bought.  This alphabet soup costs considerable time and money.
The Ethernet community is beginning to define several new form factors for 100GbE and more alphabet soup seems imminent as shown in Figure 2.  The current list of second generation form factors includes CFP2, CFP4, QSFP2 and the next generation 10X10 module that we’ll call the CXP2 (two proposed names and solutions are CXP2 and XXP[http://10x10msa.org/documents/Proposed_Module_Concepts.pdf] and the CXP could even be used with exotic cooling techniques).  The 10X10 MSA and the CFP MSA have no common members and neither MSA is expected to back down from defining these new form factors.
Figure 2: More Alphabet Soup for you?
The CXP2 already has an electrical interface defined and the optical interface will be released soon.  The CXP2 is based on the CPPI (unretimed) electrical interface defined in 802.3ba. The CFP2 and CFP4 will probably use the CAUI-4 and CPPI-4 (not official yet) electrical interface that are expected to begin being defined in the IEEE in July, 2011 in a new study group.  The new IEEE study group will also investigate the need for new optical interfaces such as 100GBASE-SR4, and 100GBASE-FR4 ( a single-mode variant for 2km?).  The single-mode optical interfaces (100GBASE-LR4 for 10km and 100GBASE-ER4 for 40km) were defined in 802.3ba with 100GBASE-SR10.  The industry will need to contribute considerable effort to get these new technologies defined and deployed.
Because of the mature nature of the 10Gbps technology of the 10X10 MSA, the cost of the 10X10 technology will remain lower cost than 100GBASE-LR4 technologies in the CFP, CFP2 or CFP4 form factors for several years.  The integrated photonics technology used in the 10X10 will keep the cost below the cost of the discrete components used the LR4 solutions until they also use integrated photonics.  The CFP2 and CFP4 will be based on new 4x25Gbps signaling that will take years to bring to market.  The alphabet soup from the 10X10 is limited to the CFP and CXP2 while the 100GBASE-LR4 soup will have CFP, CFP2, CFP4 and QSFP2 floating around for a few more years.
If you want to join the discussion on next generation 100GbE Form Factors, go to:
1)      http://10x10msa.org/documents.htm regarding presentations on the next generation form factor and join the group by emailing info@10x10msa.org.  To join the public reflector to see announcements about upcoming meetings to discuss the higher density form factor, go to http://groups.google.com/group/10x10g and click on the "Sign in and apply for membership" button.
2)      http://www.cfp-msa.org/Documents/CFP_MSA_OpGuidelines_Issue1-1.pdf to see how you can become a sponsor member to the CFP MSA.  They don’t seem to have any sponsor members or public information about their next generation form factors.
3)      http://tech.groups.yahoo.com/group/NG100GbE/ to discuss the next generation PMDs and electrical interfaces that will be defined in IEEE.
4)      The QSFP2 is being designed for 100GbE applications with 4 lanes of 25Gbps by the SFF Committee (www.sffcommittee.org) and is not designed to be backward compatible with the QSFP so that it can dissipate more power and not get too hot.
Enjoy your soup,
The 10X10 MSA

PS.  For a thorough understanding of Ethernet alphabet soup (nomenclature like 10GBASE-LX4 or 40GBASE-FR), read this white paper titled the Evolution of Ethernet Nomenclature: http://www.brocade.com/downloads/documents/technical_briefs/Ethernet_Nomenclature_GA-TB.pdf.
PSS. See how Brocade and Alcatel-Lucent plan to support the 10X10 in this NANOG presentation:

Monday, January 24, 2011

Next Gen 10x10 Module Meeting on 1/25/2011

10X10 MSA Enthusiasts,

The 10X10 MSA has seen considerable amounts of interest in the next generation module form factor meeting this Tuesday.  We will meet to discuss the next module that is between a CXP and a CFP module.  The CXP is too small to dissipate enough power for the 10X10 and the CFP is too large to give a high port density.  The 10x10 MSA is inviting interested parties to share their requirements for this next generation module.

Logistics:

When: Tuseday, January 25th from 1:30 to 4.  Please show up by 1:15 to get a seat or you may have to sit in the overflow room.

Where: 900 Alta Ave, Mountain View, CA 94087 in the Googleplex.  You will be the guest of Bikash Koley of Google and each company needs to limit physical participants to two people per company.  There will be two rooms that seat about 25 people each and will be linked with video conferencing.  The first room will most likely be limited to only one company representative since 25 companies have already shown interest.  The second overflow room will be available for the second participant.   

For those who would like to use the teleconference, please call:
Reservationless-Plus toll-free dial-in number (U.S. and Canada): (888) 878-2699
Reservationless-Plus international dial-in number:(706) 643-7329
Conference code: 9139442205
Security Passcode: 1234


Presentations for the event will be posted at: http://10x10msa.org/documents.htm

What: The agenda for the meeting is as follows:
1:30 – Introductions and greetings – Scott Kipp of Brocade
1:45 – 100GbE for Warehouse Scale Computing - Bikash Koley of Google
2:15 – Requirements for Next Generation 10x10 Modules – Scott Kipp of Brocade
2:45 – Proposed Module Concepts – David Lewis of JDSU
3:15 – Open Discussion
3:55 – Next Steps and Adjournment

Who: The 10x10 MSA has grown to 13 members including Santur, JDSU, Google, Brocade, Enablence, Cyoptics, AFOP, MRV, Oplink, Hitachi Cable America, EXFO, Huawei and Kotura.  Other companies expected to participate in the event include Molex, Infinera, Reflex Photonics, Oclaro, Applied Micro, LightCounting, Tyco Electronis, Amphenol and Broadcom.

Please RSVP me if you are not on this list.

Why: The demand for 100G applications is growing with the Internet and other digital ventures.  Cost effective, long reach solutions are needed to deliver bandwidth to billions of users.  The 10X10 is the lowest cost solution for single-mode fiber at 100G speeds.  Read more about it here: http://10x10msa.blogspot.com/2011/01/save-over-100000-with-10x10-link.html

To join the public reflector for 10X10, go to http://groups.google.com/group/10x10g and click on the "Sign in and apply for membership" button.

See or hear you soon,
Scott Kipp
10x10 MSA Chair

PS. Please forward this to interested parties.

Wednesday, January 12, 2011

Save Over $100,000 with a 10x10 Link

It might interest the reader to know that the list price for a 100GBASE-LR4 CFP module is over $100,000.  If you search google on “100GBASE-LR4 Price”, you’ll find PC Microstore sells 100GBASE-LR4 CFPs for $186,566.23 (excluding shipping) - http://www.pcmicrostore.com/JUNIPER-100GBASE-LR4-CFP-PLUGGABLE-COMPLIANT-MODULE-WITH-IEEE-8023BA/cat-p/c/p3548688.html . PCI Microstore says this is half off the list price of $373,121.27, so they offer a nice savings over what some people might have to pay. 
If you really want savings, use 10x10 modules for your 100GbE link.  The 10x10 MSA has the goal of the 10x10 module being less than half the price of the 100GBASE-LR4 CFP.  Since two modules are needed to create a link, the savings for a link would be about the cost of a 100GBASE-LR4 module or $186,568.23. 
In Table 1, the cost of a 10x10 CFP is assumed to be $1 less than half the cost of a 100GBASE-LR4 CFP ($186,566.23 / 2 – $1 = $93,282.13).  The cost of the modules for a 10km link is about $370,000 for 100GBASE-LR4 and about $190,000 for the 2 km 10x10 link.  Since these are list prices and discounts are probably available, it is safe to assume that a user can still save over $100,000 if their link is only going a few hundred meters or a couple of kilometers.
Table 1: Cost Savings with a 10x10 Link
Module Type
 Cost of Module
Number of Modules
 Module Cost per Link
100GBASE-LR4 CFPs
 $  186,566.23
2
 $       373,132.46
10x10 CFPs
 $    93,282.12*
2
 $       186,564.23


Savings
 $       186,568.23


* This is a calculated price that is $1 less than half the advertised price of the 100GBASE-LR4 module that is referenced above.  This is not a suggested sales price for the module which is likely less.
It is easy to see the capital expenditure (Capex) savings from using the 10x10 solution over the 100GBASE-LR4 solution if the user doesn’t need to go over 2 km.  This blog post only compares the cost of replacing the 100GBASE-LR4 CFP with a 10x10CFP and does not discuss the additional costs of the 100GbE blade, switch or single-mode fibers which would be the same for either implementation.
Google has shown that the cost of the 10x10 CFP is expected to be about the price of a 10GBASE-LR XFP in 2012 on a Gbps basis, while the 100GBASE-LR4 solution is still over 6 times as expensive.  This Google presentation (http://10x10msa.org/documents.htm) was given at ECOC 2010 and showed how the cost of the 10x10 solution (called LR10 in the presentation) will reach price/bandwidth parity with the 10GBASE-LR XFP in 2012.  The other solutions, specifically 100GBASE-LR4, will always be more expensive than 10x10 because of low volumes and its inherent cost structures. 

The 10x10 solution also saves with regards to operating expenses (Opex).  Since the 10x10 consumes about 14 Watts/module versus 20W/module for a 100GBASE-LR4 solution.  A link will save about 12 Watts or 105 kWh/year.  When you consider the power to cool the modules, the Opex savings add up fast.
The 10x10 CFP solution is designed to be the most cost effective single-mode solution in the market.   This post shows how the initial price lead of the 10x10 puts it so far ahead of the competition that they will never catch up.  After saving capex, the 10x10 will also save you opex.  What more could you ask for?

Sunday, January 9, 2011

Why is the 10x10 Standard Needed?

The 10x10 optical standard is fulfilling the need for intermediate link distances of 100GbE within the data center.  The IEEE has defined a number of links for 40GbE and 100GbE as shown in Table 1, but the difference between 100 meter links and 10,000 meter links is too big when you consider that a 100GBASE-LR4 solution can be over 10 times more expensive than a 100GBASE-SR10 link.  The 10x10 solution typically costs less than half the price of the 100GBASE-LR4 solution and bridges the gap between cost effective 100 meter links and very expensive 10km links outside of the data center.  The 10x10 modules also consumes about 12 Watts less power on a link than 100GBASE-LR4.  The 2km distance of the 10x10 offers a cost effective and efficient solution for large data centers that need to go farther than 100 meters but less than 10km
Table 1: 40 and 100GbE Variants
Backplane
CR
SR
FR
10x10
LR
ER
40GbE
1m
7m
100m
2km
N/A
10km
N/A
100GbE
Being defined
7m
100m
N/A
2 km
10km
40km

The 802.3ba project (http://www.ieee802.org/3/ba/index.html) that defined 40GbE and 100GbE limited the scope of the project so that the project could be completed in about 4 years.  The Higher Speed Study Group (HSSG) started in July 2006 and the 40 and 100GbE standard was finished in June 2010.  To not delay the standard, they only defined 100 meter, 10 km and 40 km fiber optic solutions.  If more distances were added between 100 meters and 10km, then the standard would have taken longer and the Ethernet community was hesitant to define other links and delay the release of the standard.
The first follow on project to 802.3ba related to 40 and 100GbE was the 802.3bg task force that defined a 2km link as 40GBASE-FR (http://www.ieee802.org/3/bg/index.html).  In November 2010 at the hotel with the IEEE meetings, the Next Gen 100GbE Optical PMD Discussions had a straw poll to start a Call For Interest (CFI) specifically for a 2 km solution that could support the 10x10 solution. The motion was voted down by a fact of 4 (5 yes to 21 no).   Members of the IEEE were not very interested in defining a 100GbE link at 2km in a separate standard, so the 10x10 MSA was created to fill that need.  It was clear to members of the 10x10 MSA that several man-years of effort and a couple calendar years of time would be needed to create the IEEE standard for 100GbE at 2 km.  Instead of going through this painful process, the 10x10 MSA was formed. 
The 10x10 solutions is designed to meet the needs of users who need to go beyond 100 meters but less than 2 km.  Many data centers have link requirements beyond 100 meters, but don’t need to go much more than a few hundred meters.  The 10km solutions for these applications is overkill because the 10x10 solution can meet the link requirements at less than half the cost of 100GBASE-LR4 and about 70% of the power (14W for 10x10 vs 20W for 100GBASE-LR4).  Since the 10x10 is CFP compatible and can fit in the same port as 100GBASE-LR4, customers will see the benefits of the 10x10 over 100GBASE-LR4 for link distances over 100 meters but under 2 km.