[optical-networks] Status?

[Joe Touch]

FWIW - I just finished typing an answer to this; here it is for the group:

At this point, we don't have sufficient information to determine for 
sure how useful the 5 of 8 (or K of N) correlator will be. Getting 
statistics has proven more complicated than expected.

That said, there are a few key points, regarding the OFC paper, that 
might be worth focusing on:

     1. the need for a partial solution to route lookup indicates
         that looking at subsets of the bits might be useful
         (i.e., the design is motivated by a possible need)

     2. the correlator can be used for all sorts of triage, not
     just routing
         i.e., to load-balance traffic. in this case, the
         correlator configuration isn't at the mercy of the
         results of an external routing algorithm; the need
         to look at small sets of bits can be used BY the
         load-balancer ensure that correlator-compatible sets
         are selected

     3. there have been several important results from your design,
     not so much from a pure optical perspective, but on their
     impact on the algorithms used to configure correlators:

         a) 5 bits is OK, 10 is not (?)
             limiting the number of bits then limits
             the number of combinations of bits that the
             algorithm needs to examine.

         b) 0 bits are 2x (?) easier to examine than 1 bits
             if we're doing load balancing, then we want
             to use patterns that prefer 0's over 1's to
             split traffic

Some useful stats we have so far, BTW:

     measured at a core-edge router
             104,400 routes
             18 interfaces,
                 3 carry 93% of the routes
                 4 (3+one more) carry 99% of the routes
             most of the routes are into the core
             most of the routing difference is load balancing
                 this is typical for core-edge, but
                 not core-middle routers

I don't know if any of the stats above are useful to justify the design 
at this point, but feel free to use them if they are.

Michelle C. Hauer wrote:

> I was just wondering how things are going over there.  I am working on
> the figures for the paper right now and then I will start editing the
> content.  I'm going to put Dr. Bannister's intro paragraph into the
> short "algorithm" section of the paper.  Is there some sort of figure
> (table, etc.) that can complement this?  Or do you have any results from
> Dr. Touch's work with the actual routing tables that I can show?  We are
> very constrained on room so I might have to shorten whatever you give
> me, but we'll see how it goes.  Since this is an optics conference, I
> need to make sure the optics novelty really comes through and that the
> networks part is interesting and understandable.
> 
> I'll be working on this and my other two papers between now and the
> deadline (except for a little sleep tonight), so I will be in constant
> touch by email or phone (213-740-0024).  When I get far enough along,
> I'll send out a copy of the paper so you can see how it is going.


I just tried to call about 10 minutes ago; call me if I can help clarify 
any of this... (310-448-9151)

Thanks,

Joe