6.7 Meter Telescope To Capture 30 Terabytes Per Night

Lumenary7204 writes "The Register has a story about the Large Synoptic Survey Telescope, a project to build a 6.7 meter effective-diameter ground-based telescope that will be used to map some of the faintest objects in the night sky. Jeff Kantor, the LSST Project Data Manager, indicates that the telescope should be in operation by 2016, will generate around 30 terabytes of data per night, and will 'open a movie-like window on objects that change or move on rapid timescales: exploding supernovae, potentially hazardous near-Earth asteroids, and distant Kuiper Belt Objects.' The end result will be a 150 petabyte database containing one of the most detailed surveys of the universe ever undertaken by a ground-based telescope. The telescope's 8.4 meter mirror blank was recently unveiled at the University of Arizona's Mirror Lab in Tucson."

Comments

This is why...

[ShadowFalls]

It not being in space might have something to do with the amount of data it would have to transmit and the speed limitations... Besides, you can't replace Hubble, its impossible to exactly replicate that many technical difficulties...

30TB raw?

[Jah-Wren Ryel]

When I worked at the CFHT a few decades ago, they had a bunch of "data reduction" algorithms they ran on each night's run that reduced the amount of data they needed to store by at least a factor of 10.

Re:30TB raw?

[$RANDOMLUSER]

What, like storing the year as two digits, that sort of thing?

Re:30TB raw?

[ShadowFalls]

No, removing any data that could prove the possibility of Extraterrestrial life :P

Re:30TB raw?

[mrsquid0]

Back when I was using CFHT there was no high-pass filtering done on the data. That would change the noise properties of the data, which could render the data useless for certain types of analysis. The big space savings were done using lossless data compression. Depending on the type of data one can reduce the disk space required by up to about 90%. A second space-saving technique was to combine calibration data, such as bias frames and flats. In many cases combined calibration data is just as good as the individual frames, and in some cases better. Roughly half of the data collected each night is calibration data, so this can result in a big saving in space. I have not used CFHT since the 1990s, so I have no idea how they deal with their data now.

Any colour you like so long...

[Harold Halloway]

30 Terabytes, consisting mainly of #000000.

Re:Any colour you like so long...

[NixieBunny]

I used to think that until I started to work for astronomers. They actually take photos of noise, and then add noisy images together ("stacking") until pictures of interesting faraway things emerge from the noise. That's why a nightly sky survey is so useful - you can add together a few months of images and see stuff that you would never have seen in a single image.

it will run on MySQL

[datacharmer]

This project was presented at the MySQL Users Conference 2008 in a dedicated talk and a keynote.

The storage will be organized in clusters based on MySQL databases.

Astronomy, Petabytes, and MySQL

The Science and Fiction of Petascale Analytics

Sounds like a big number, but..

[bertok]

30 TB per night sounds like a lot, but 1.5 TB drives are about AUD 350 each, retail. By 2016, I'd expect vendors to have released at least a 10 TB hard drive at that price point, and I wouldn't be surprised if we're using 30 to 50 TB drives.

So it all boils down to about $1000 per night of operation, or about $350K per year. Not exactly expensive for a science project. A single mars mission costs about $300M, but this telescope would generate more discoveries. That's not even considering that storage costs would continue to drop over the lifetime of the telescope, so the eventual total cost may be less than $100K per year. That's the salary of just one person!

Re:Sounds like a big number, but..

[dwater]

The cost of the storage might be reasonable, but what about the performance aspect? 30TB per night sounds like a lot to store in one night...being generous and calling a night 12 hours - I'm probably wrong, but I make that 43200 seconds which is 694 MB/s. Without looking up any performance stats for hard drives, that sounds fairly easily attainable (too).

Re:Why not...

[Nyeerrmm]

The basic problem is that a 6.4 meter aperture can't fit in a launch vehicle, Ares V is only to be 5.5 meters.

Hubble was built at the diameter it was (2.4 meter) because thats about the maximum you could build of a stiff mirror that held its shape well enough through launch to remain optically sound on orbit. When you require 10-nm level precision, it takes a hefty structure to keep things that stiff.

In order to go bigger, the methods they're using for James Webb manage to double the aperture while halving the weight. The way they do this is using active controls and sensors to correct errors rather than rely on avoiding all errors. But looking at James Webb, you'll notice it focuses on IR which is very hard to observe from Earth, while no optical band concept is out there. This is because the new, big Earth-bound scopes use adaptive optics to eliminate seeing errors (the variations in the atmosphere that Hubble avoids), and get potentially better images than Hubble since larger mirrors can be used.

Of course, if the money shows up, there are other advantages to having a space-based observatory, particularly access time and not having to worry about the effectiveness of the adaptive elements, so I'm sure we'll see a proper Hubble replacement eventually, but it's certainly not as critical for scientific progress as some might think.

Re:Harddrive space doubles every year...

[lysergic.acid]

we might be reaching the physical (or practical) limit of data density for hard disk platters, but we'll probably just move to new technologies. it's very unlikely that magnetic disk drives are the pinnacle of data storage technology. there are probably more efficient storage mediums in the works already.

i don't know what can currently match magnetic disks drives in terms of IO speed, but holographic storage shows a lot of promise. in theory, holographic storage can read/write millions of bits of data in parallel rather than one bit at a time as with conventional optical or magnetic media. the theoretical limit of holographic storage density is tens of Tb (terabits) per cm^3. and already commercial industries have achieved 500 Gb per square-inch (about 5x the density achievable on magnetic hard disks).

holograms also have some very interesting properties that may or may not transfer to digital data storage. for instance, if you record a hologram of a 3D object onto a photographic plate, you can in essence reproduce a 3D image of the whole object with any piece of that photographic plate. Wikipedia explains this phenomenon thusly:

Since each point in the hologram contains light from the whole of the original scene, the whole scene can, in principle, be re-constructed from an arbitrarily small part of the hologram. To demonstrate this concept, the hologram can be broken into small pieces and the entire object can still be seen from each small piece. If one envisions the hologram as a "window" on the object, then each small piece of hologram is just a part of the window from which it can still be viewed, even if the rest of the window is blocked off.

since holographic data storage also uses optical interference patterns to store information, i guess it's possible that this phenomenon would also transfer over, though it might not since we're talking about digital data in this case rather than analog data. with analog data, losing a part of the interference pattern simply reduces the resolution of the holographic image, though it remains whole. with digital data, that loss of resolution could simply corrupt the data. but i don't know, i'm not a holography expert.

Nits and Grins

[martyb]

From TFA title: (emphasis added)

6.7 Meter Telescope To Capture 30 Terabytes Per Night

<nit>
That's 6.7 Meter effective diameter Telescope. The primary mirror has a diameter of 8.4m but the tertiary mirror (5.2m diameter) sits right in the middle of the primary, so its area needs to be subtracted from the primary. The area of the primary is pi*(8.4/2)^2 which is 55.4m^2 and the area of the tertiary is pi*(5.2/2)^2 which is 21.2m^2; a single mirror of that area would have a diameter of about 6.7m.
</nit>

6.7 Meter Telescope To Capture 30 Terabytes Per Night

<grin>
Hey!! I thought information wanted to be free! And here they plan to go off and capture 30 TERAbytes? Each night? OMG!!!!11Eleventy!! Say it ain't so!!
</grin>

Re:A waste

[resignator]

It would be far less cost effective to upgrade Hubble than to build the LSST. Shuttle launches cost a TON of money not to mention you are risking astronauts lives to try and upgrade something we CAN build better on Earth. Hubble has had optical problems in the past and the 2 flights to repair it cost about one billion dollars. Imagine what a complete rebuild would require. "Making the most advanced telescope, ON THE GROUND, seems like an oxymoron to me." And the reason it is considered such an advanced telescope: http://en.wikipedia.org/wiki/Adaptive_optics