The Missing Link: Dedicated End-to-End 10Gbps Optical Lightpaths for Clusters, Grids, and Clouds

[Venessa Miemis]

[Over the past several weeks there has been a lot of exciting
developments in the R&E network community at the Spring Internet 2
meeting and more recently at the Terena meeting in Prague about new
Internet architectures built around GOLEs – GLIF Open Lightpath
Exchanges. There is a growing consensus amongst most of the major R&E
networks that we need to move away from traditional hierarchical
networks to GOLEs enabling the direct peering of networks,
institutions and/or researchers. This is primarily being driven by the
data demands of global collaborative research. GOLEs enable a policy
free interconnection with no bandwidth constraints or blocking between
the connecting parties and therefore research is not constrained by
policy or bandwidth issues as on traditional networks. GOLEs are also
important for the R&E networks in the new research intensive nations
like Brazil, China, South Africa, Korea, etc. In the future networks
and researchers of these nations will be far less dependent on having
to transit intermediary networks by peering directly with networks or
institutions in designated host nations. The CANARIE UCLP (User
Controlled LightPaths) and more recently Internet 2’s OS3E service
are intended to allow institutions, or even researchers to establish
their own private networks for specific VOs or communities of interest
who can interconnect at these GOLEs.

There is no question that GOLEs are going to challenge traditional
business models for R&E networking, but the first priority must be to
enable the needs of the researchers themselves in the exponential
growth of data driven science. Here are 2 excellent slide
presentations explaining this explosion of data driven science and why
we need GOLES – BSA]

The Missing Link: Dedicated End-to-End 10Gbps Optical Lightpaths for
Clusters, Grids, and Clouds

Dr Larry Smarr

http://lsmarr.calit2.net/presentations?slideshow=8084045

Dynamically Provisioned Networks as a Substrate for Science

David Foster CERN

https://lhcone.web.cern.ch/node/75

[Colin Hawkett]

Ok so I had no idea what the article as talking about :) I looked it up and actually had some difficulty getting a decent explanation - this article seemed to do an ok job of explaining it -

http://www.ces.net/doc/press/2010/pr101208.html

And I got this from it (could be totally wrong) -

1. The 'peers' must be connected by fibre optic cables, which carry information encoded as 'lambdas'.
2. Each peer is an exchange, which is able to join two lambdas together and make the two endpoints seem connected directly to each other (knock out the exchange and the connection would fail).

So, it sounds like universities running fibre between themselves - i.e. creating their own network, rather than travelling on the current net. The topology seems very static (defined by real fibre connections). The existence of this auxiliary network would be good thing from the perspective of government control of internet connectivity, and from an alternate infrastructure perspective, but I didn't get a clear picture how independent this is proposed to be. I guess it kind of goes back to the roots of the net where it used to be just R&E running the thing in a lot of places. Perhaps this is the evolutionary cycle. I'd reckon this is the sort of scale and type of uptake required to deliver the beginnings of a next gen infrastructure. Getting the money for it without private investment would be a big achievement - definitely would be a great thing to have running. The fibre networks they have already laid down in the US for the project seem impressive.

--------

Here is a tangentially related article talking about the creation of a dedicated distributed system to handle the data and processing volumes from Square Kilometer Array radio telescope project. They're talking about the project generating up to an exabyte of data per day when fully operational.

http://www.computerworld.com.au/article/387097/exclusive_ska_bid_looks_skynet_computing/

Just for fun, if you're wondering how much data an exabyte is - http://www.readwriteweb.com/solution-series/2011/05/how-big-is-a-yottabyte-infographic.php

[Marcos]

On Wed, May 25, 2011 at 8:03 AM, Colin Hawkett <haw...@gmail.com> wrote:

Ok so I had no idea what the article as talking about :) I looked it up and actually had some difficulty getting a decent explanation - this article seemed to do an ok job of explaining it -

My guess, is that's it's an attempt to make a *fractal* network topology (physically) rather than a flat graph at layer 0.  In theory, since physical geography (and user base) is fairly stable (correlated with population) this should allow a more optimized network with fewer nodes and costs (probably by an order of magnitude or several).  Such a topology could be very relevant to making the open mesh internet...

just a guess

marcos