Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [HSSG] Reach Objectives: distribution of reaches



See in line...

> -----Original Message-----
> From: Drew Perkins [mailto:dperkins@xxxxxxxxxxxx]
> Sent: 23 August 2006 08:53
> To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
> Subject: Re: [HSSG] Reach Objectives
> 
> Why? Because the debate has been had repeatedly for every 
> generation of technology that has come and gone and will 
> probably be had for every generation that is yet to come. The 
> underlying technology changes and that will change the answer 
> in terms of where the breakpoints lie between different 
> technologies.

True.

> Another big thing that changes is the set of 
> people in the room and different groups will always come up 
> with different answers based on their different perspectives. 
> But one thing that I believe changes very little from 
> generation to generation is the reach between different 
> boxes.

Not true.  802.3ae like previous generations of Ethernet was primarily about campus and general-purpose wiring, with an attempt to connect to the WAN.  In the context of HSSG, in my opinion, in the nearer term, the very fat routes for which 10G*Cl.43LAG are insufficient are WITHIN supercomputer rooms and similar.  In the future, the number of fat routes going to another supercomputer room in another city may rise above our threshold of interest.  I just don't see a significant number of campuses with TWO tightly coupled top-end supercomputers or monster routers on the one campus, any time soon.  So I see HSSG's primary interest as truly short links - forget the campus!

> My comments are mostly influenced by my perspective on 
> telecommunications applications, but I suspect the same thing 
> mostly applies to datacenters too. The shape of the histogram 
> of the number of links plotted vs. the length of those links 
> probably stays roughly constant. Where the breakpoints are of 
> greatest cost effectiveness between different fundamental 
> technologies will move depending on bit rate, but to the 
> first order the number of PMDs probably remains roughly equal.

Looked at that way, the distribution is very broad without very clear break points.  The 802.3ae study of all this was unsatisfactory because even after all the study, it seemed like a judgement call (apart from the 40 km choice for the then repeater hut spacing).  By giving into a natural wish for "just a few more meters" when the technology wasn't developed enough for a consensus understanding of the costs and engineering detail, we ended up with a PMD set that's simply poor for the equipment room: too power hungry.

And if one then filters the distribution of links to find the distribution of fat (super-sized?) routes, I believe the convex distribution becomes more of a bathtub.

Piers
 
> My thoughts are further influenced by my expectation that for 
> multi-km reaches we'll likely end up with Nx10G (or N/2x20G) 
> WDM PMDs. The smaller the N and the larger the bit rate, the 
> more the breakpoints are likely to move around.
>  
> Drew
> _____________________________
>  
> Drew Perkins
> Chief Technology Officer
> Infinera Corporation
> 1322 Bordeaux Drive
> Sunnyvale, CA  94089
>  
> Phone:  408-572-5308
> Cell:       408-666-1686
> Fax:        408-904-4644
> Email:    dperkins@xxxxxxxxxxxx
> WWW :  http://www.infinera.com
>  
> _____________________________
>  
> 
> 
> 
> From: Brad Booth [mailto:bbooth@xxxxxxxxxxxxx] 
> Sent: Tuesday, August 22, 2006 10:56 AM
> To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
> Subject: Re: [HSSG] Reach Objectives
>  
> Excuse my ignorance, but why?  When 10G started, this was an 
> important debate.  If they had stuck with the number of PMDs 
> in 802.3z, then there would have only been two PMDs.  The 
> study group needs to define the markets that it wants to 
> satisfy, and only then when the study group becomes a task 
> force can the group decide on the actual number of PMDs required.
>  
> Thanks,
> Brad
>  
> 
> 
> 
> From: Drew Perkins [mailto:dperkins@xxxxxxxxxxxx] 
> Sent: Tuesday, August 22, 2006 10:45 AM
> To: Brad Booth; STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
> Subject: RE: [HSSG] Reach Objectives
> I have a simple proposal to save everyone a lot of time. 
> Let's make the baseline plan to stick with exactly the same 
> number of PMDs with exactly the same reaches as 10GbE. They 
> have all been developed to fit someone's requirements. We 
> should have a very high hurdle to eliminate any of these or 
> add to them.
>  
> Drew
> _____________________________
>  
> Drew Perkins
> Chief Technology Officer
> Infinera Corporation
> 1322 Bordeaux Drive
> Sunnyvale, CA  94089
>  
> Phone:  408-572-5308
> Cell:       408-666-1686
> Fax:        408-904-4644
> Email:    dperkins@xxxxxxxxxxxx
> WWW :  http://www.infinera.com
>  
>  
> _____________________________
>  
> 
> 
> 
> From: Brad Booth [mailto:bbooth@xxxxxxxxxxxxx] 
> Sent: Tuesday, August 22, 2006 10:19 AM
> To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
> Subject: Re: [HSSG] Reach Objectives
>  
> The 10km vs. 2km debate occurred in 802.3ae.  It was 
> determined that the solution for 2km was virtually the same 
> as that for 10km; therefore, the task force felt it would be 
> better to select only one of those reaches and 10km was it.  
> If there is a desire for 2km and 10km reaches in HSSG, then 
> in my humble opinion, those wishing to support will need to 
> bring justification forward as to why the study group should 
> consider possibly increasing the number of PHYs.
>  
> Cheers,
> Brad
>  
> 
> 
> 
> From: Frank Chang [mailto:ychang@xxxxxxxxxxx] 
> Sent: Tuesday, August 22, 2006 9:19 AM
> To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
> Subject: Re: [HSSG] Reach Objectives
> John/Petar;
>  
> I agree datacenter for up to 100m. How about 10GigE LR 
> disctance (10km) or campus 2km discussed earlier? 
>  
> Frank
> -----Original Message-----
> From: Petar Pepeljugoski [mailto:petarp@xxxxxxxxxx]
> Sent: Tuesday, August 22, 2006 6:02 AM
> To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
> Subject: Re: [HSSG] Reach Objectives
> 
> John, 
> 
> I believe that the Data Center distances should be 100m and 300m. 
> 
> Petar Pepeljugoski
> IBM Research
> P.O.Box 218 (mail)
> 1101 Kitchawan Road, Rte. 134 (shipping)
> Yorktown Heights, NY 10598
> 
> e-mail: petarp@xxxxxxxxxx
> phone: (914)-945-3761
> fax:        (914)-945-4134
> John DAmbrosia <jdambrosia@xxxxxxxxxxxxxxxxxxx> 
> 08/22/2006 12:37 AM 
> Please respond to
> John DAmbrosia <jdambrosia@xxxxxxxxxxxxxxxxxxx>
> ToSTDS-802-3-HSSG@xxxxxxxxxxxxxxxxx 
> cc 
> Subject[HSSG] Reach Objectives
> 
>  
>   
> 
> 
> 
> 
> 
> All, 
> We have had some conversation on the reflector regarding 
> reach objectives.  Summarizing what has been discussed on the 
> reflector I see the following - 
>   
> Reach Objectives 
> Long-Haul   --> 100+ km 
> Metro       --> 10+ km 
> Data Center --> 50m & 300m 
>   
> Data Center Reach Segregation 
> Intra-rack 
> Inter-rack 
> Horizontal runs 
> Vertical risers 
>   
> Use this data to identify a single low-cost solution that 
> would address a couple of the reach objectives 
>   
> Other Areas 
> During the course of the CFI there were individuals who 
> wanted Backplane Applications kept in for consideration, but 
> I have not heard any further input in this area.  Are there 
> still individuals who wish to propose Backplane as an objective? 
>   
> John 
>   
>