Re: GENI discussion

[David Farber <dave () farber net>]

Begin forwarded message:

From: Bob Frankston <bob37-2 () bobf frankston com>
Date: May 22, 2007 4:50:28 PM EDT
To: dave () farber net, ip () v2 listbox com
Cc: "'Ellen Witt Zegura'" <ewz () cc gatech edu>, "'Scott Shenker'"  
<shenker () icsi berkeley edu>, "'Lazowska'"  
<lazowska () cs washington edu>, dpreed () reed com
Subject: RE: [IP] GENI discussion

I’ll admit that much of this is an expression of my own frustrations  
as I try to explain why Internet Inc is the wrong model just as I try  
to explain why broadband is the anti-Internet. I see GENI as a  
symptom of incrementalism even though it’s just a funding mechanism  
for research. I realize that it is not really supposed to be the next  
Internet but nonetheless it is filling that role by being a focus of  
Internet research. I don't blame the sponsors for the hype but it's  
still useful to take a look at it as a proxy for the idea that we  
should look at the network architecture rather than at how we use  
connectivity. I’m also biased towards solutions that empower the  
individual rather than institution be they corporations or  
governmental agencies.



If there indeed equivalent “from the edge” efforts I would like to  
know about them. For now I’m treating GENI is the face of Internet  
research whether or not it’s meant to be. It’s a useful foil for  
contrasting the network-centric approach to alternative approaches.



If GENI is indeed "a facility for experimenting with new Internet  
architectures" then the absence of the null case is telling. This is  
not necessarily the fault of those sponsoring GENI or other such  
efforts except as a reminder of the lack of funding for approaches  
that aren't as well-defined.



I've been arguing that the Internet has been defined by the end-to- 
end argument as a constraint -- that is, the inability to depend on  
any particular architecture. Attempts to improve the architecture  
with smart protocols like multi-casting and to extend it with V6 have  
been problematic while P2P approaches have met with wide spread  
adoption. These P2P efforts are the real inheritors of the Internet's  
End-to-End constraint.



What most concerns me is the seeming lack of this defining constraint  
in the research efforts. It leads to efforts to solve complex  
problems like phishing and scaling in sterile environments that can't  
exhibit these problems. There seems to be a presumption that issues  
like security can be addressed in a physical architecture of a network.



I cited the light switch/fixture problem because it seems so trivial  
yet demands a solution that does not depend on the network as a  
solution-provider. All we can presume is two consenting devices that  
might be able to exchange messages. To be more precise one can indeed  
assume a network of services but what makes this problem so  
interesting is that it arises because I have not permitted myself to  
take advantage of the overwhelming temptation to rely on network  
services and thus we can decouple the system elements. Such solutions  
are not only able to exhibit Moore's law effects they are also more  
stable.



My own "research" consisted of trying to do projects such as home  
control (and, while at Microsoft, working with Honeywell and Intel to  
(unintentionally) learn what doesn't work). It's the same kind of  
constraint that I used for home networking (no installers or service  
providers) which arose from my experience with personal computing. I  
remember efforts like Project Athena failing to "get" personal  
computing.



I don't think I'm too cynical to observe that well-defined  
experiments on a network test bed are very appealing to those who  
need to have fundable projects with deliverables. Congressional  
scrutiny of funding almost requires local justification of each step  
while efforts to improve connectivity at the edge are too-easily  
treated as threatening by those who associate P2P with a loss of  
control.



The question then is where is the effort to solve problems in making  
use of whatever transport is available to do our own networking and  
to create solutions at the edge. Problems like phishing cannot be  
solved inside the network nor can they be "solved" in a closed sense.  
I also argue that efforts to solve scaling problems in a test bed  
fall into what I consider the trap of trying to solve a problem  
instead of seeing the opportunity in solutions already available.



Compositing local connectivity when we can assume abundant capacity  
is very different from presuming scarcity with a single entity can  
manage the complex routing through chokepoints. Why aren't we asking  
why we can't take advantage of the full capacity of the available  
fiber (and other transports) rather accepting this synthetic scarcity  
and reveling in the complex algorithms necessary to work around it?



How do we support projects that aren't as amenable to closed  
solutions but are still vital? This is a nontrivial question and,  
perhaps, we have to rely on by-product of the effective marketplaces.  
This is why individuals can create P2P solutions without  
organizational funding but such funding helps in doing professional  
solutions as we’ve seen with Skype and from Firefox’s challenge in  
scaling the effort. That leaves a lot of problems, like phishing,  
seemingly orphaned because it’s not as amenable to solutions in  
isolation.





-----Original Message-----
From: David Farber [mailto:dave () farber net]
Sent: Tuesday, May 22, 2007 07:55
To: ip () v2 listbox com
Subject: [IP] GENI discussion







Begin forwarded message:



From: "Zegura, Ellen Witte" <ewz () cc gatech edu>

Date: May 22, 2007 3:33:53 AM EDT

To: ip () v2 listbox com

Cc: dave () farber net, Scott Shenker <shenker () icsi berkeley edu>, Ed

Lazowska <lazowska () cs washington edu>

Subject: GENI discussion



I’ve seen the IP thread about GENI that was prompted by yesterday’s

announcement that BBN will be the GENI Project Office (GPO).  Along

with Scott Shenker, I am helping lead the GENI Science Council (GSC),

the group that represents the research community that will make use

of the GENI facility.







I wanted to respond to a few parts of Bob’s message.  Primarily, I

wanted to make clear that GENI is a facility for experimenting with

new Internet architectures, not a proposal to BE the new

architecture.  The reason for slicing is to allow multiple

experiments to run on the same physical resources at the same time.

The slices are managed so that experimenters don’t step on one

another during the testing phase.  There are other funding programs

in NSF that will fund the research on alternative architectures and

ideas that can be tested in the GENI facility.  These include the

FIND program, and many other networking and distributed systems

programs.  Of course, one could propose that GENI itself be the new

architecture – and surely some will -- and then folks like Bob and

others could reasonably debate the wisdom of that.  We’re not there

yet, though, and its important to keep the distinction between GENI

as a facility and GENI as a proposed new architecture.







With respect to the suggestion that money would be better spent on

research that improves our understanding of how to operate well from

the edge, I guess I view that as complementary.  GENI is likely to

support experiments that modify edges and use a standard IPv4 or v6

core, perhaps with instrumentation inside the network that would

allow more insight into what helps edge-controlled apps work better.







Ellen















-------------------------------------------

Archives: http://v2.listbox.com/member/archive/247/=now

RSS Feed: http://v2.listbox.com/member/archive/rss/247/

Powered by Listbox: http://www.listbox.com



-------------------------------------------
Archives: http://v2.listbox.com/member/archive/247/=now
RSS Feed: http://v2.listbox.com/member/archive/rss/247/
Powered by Listbox: http://www.listbox.com