Slashdot Mirror
Researcher Only High Bandwidth Network
Icarus1919 writes "A brand-new 10 gigabit per second per user optical fiber network is now available to researchers in the U.S. (compared to Internet2, which offers only 10 gigabits of bandwidth total, regardless of the number of users). The National Lambda Rail, as it is known, is named for the 40 different wavelengths of light it uses to send data within the fiber network. In the past, researchers have complained about the relatively (relative when you're dealing with terabytes of data) small bandwidth they can access to send data, and the addition of the NLR will most likely be a boon to research."
Exactly. Every guarantee has an equal and opposite refusal. In this case they have only 40 lambdas, so they can only support 40 users at a time.
They're both 10Gbps, but the difference is that this new one is 10Gbps per *user,* while Internet2 is 10Gbps for *all* users. That's a rather significant difference, I suspect.
Bít, zabít, jen proto, ze su liska!
10Gb/s = 1TB transfered in 800 seconds, or 13 minutes 20 seconds. 10Gb/s per user is far far greater than 10Gb aggrigate.
The Doormat
If you're not outraged, then you're not paying attention.
By "user" they mean "institution connected to our network" not "individual person". As the previous reply said, they're limited to 40 wdm channels, one per user. To put this in perspective (from wikipedia):
Anyone know what the shannon limit for single mode fiber is?
-jim
More information is available at nlr.net including a network map. The first link that went up was between Chicago and the Pittsburgh Supercomputer Center which is run by CMU (where I go). The only problem I think we only have 1 (maybe 2) gigabit links to them, so the bandwidth isn't available onto campus.
While some tests have been run with simple off-the-shelf-hardware, in most cases the 10Gbps connections will be shared amongst a cluster of machines or a high-end machine.
For some relevant projects actually using NLR right now:
TeraGrid
OptIPuter
It is not really 10 Gbps per user. New Scientist got that wrong. It's a fiber-optic infrastructure capable with 40 lambdas max. A lambda is a wavelength of light. They use DWDM to split the light on the fiber into 40 lambdas. One lamdda = 10 Gbps.
Now, a single user can, for a period of time, get a whole lambda for himself for a particular application. That's a big deal for researchers. But don't think that everyone at an NLR connected institution automatically has a 10 Gbps link to everyone else on NLR. Most of the users, at best, probably have 1 Gbps ethernet to their desktop. This isn't for browsing the web and playing Doom. It's to connect huge data stores in San Diego to supercomputers in Pittsburgh.
Internet2 is transparent between hosts that are both on Internet2. College students across the country are using I2 for political discussion, gaming, and pr0n right now, and never even knowing it.
Give me Classic Slashdot or give me death!
The point of the lambda rail system is not necessarily the bandwidth. Its that there is no "routing" involved on the base network in the traditional sense. High bandwidth combined with low latency is the goal.
--
WHO ATE MY BREAKFAST PANTS?
Last week we celebrated
the 50th anniversary of CERN (European Laboratory for High Energy Physics).
There was an "Open Day" organized (visits to tens of interesting places at CERN) and the Computing Center was one of the most popular places.
I was honoured to be one of the guides there.
The LHC will produce proton-proton collisions 40 million times per second.
Even after a first pre-selection of the events, there will be around 100 "potentially interesting" events per second that have to be recorded for further (so called "off-line") analysis.
There will be 4 detectors running on LHC (ATLAS, CMS,LHCb and ALICE).
The size of the data describing a typical "event" (collision) depends on the detector, but it is of order of Megabytes.
Assuming 1MB/event/detector that gives
1MB*100events/s*4 detectors*86400s/day=
34560000 MBytes of data produced every day.
That is almost 35 Terabytes of data per day (24 hours).
We expect to have some 10 Petabytes of data per year. And we hope to run the LHC for some 20 years.
Obviously, this amount of data cannot be stored/analysed locally; it will be therefore transmitted to scientific institutes worldwide.
You may have a look at the presentation:
[PPT]
PDF
The answer to computing power is the Grid: more information at http://gridcafe.org/
cheers,
Piotr Golonka
CERN IT/CO
I have not enough data to compare the performance of this network with the findings of the Law.
You will be surprised to see that it has a shorter cycle.
Well, there's a useful pamphlet (PDF, 830 kB) on the National LambdaRail website.
It states that there will be a solicitation/application process that will peer review project proposals. The intent is for NLR to be used both as a tool for other research, and also for research into networking technology (both protocols and hardware).
To answer your question, you need to convince their scientists that you have an interesting project proposal, and you probably need to be "affiliated with" a big chunk of grant money.
~Idarubicin