![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Entry tags:
Space is big, and needs a lot of storage
I was teeth-grittingly determined to make it to tonight's ACS meeting, so while I achieved nothing else today, I did get there to hear about the Square Kilometre Array and sundry side organisations and projects.
The gist of the talk was that (a) space is big, (b) we're going to be collecting an enormous amount of data in a relatively short time, and (c) all that data has to be stored, shifted, accessed, analysed, and accessed again. The numbers were insanely large; as the branch chair said (approximately): "We're used to dealing with zeroes and ones, but you've got a lot more zeroes than we're used to."
Since the project is partly a public relations excercise to convince some committee that we need to build this thing in Australia (rather than South Africa), a lot of effort has gone into the promotional materials. The speaker was geeky but accomplished, and there were bonus posters and glossy conference packs.
I myself am not a numbers girl. Tetra-, petra- and exo-flops don't say much more to me than "A lot". Where I can translate that into human and logistical terms, it's more bandwidth than the sea, and the same volume of data as internet traffic in a week. All this incoming into server farms so vast that even under the best conditions, failures are predicted at one per day. The practicalities of getting so much data to Perth from the Murchison before the cabling is done mean that they're seriously considering using utes once a week to collect that week's data and set up the new drives. The projections of computer grunt required run right along to the top edge of projections for the next 10 years. This with be FAST.
We were also promised 10 gague fencing wire and gaffer tape
I was impressed and saw through the vast quantity of the numbers to the infrastructure and people required to keep all this running. Yes, I'm attached to the human element in large enterprises :-)
I'm surprised they haven't started calling it SkyNet yet, and borrowing all this vast computing power to experiment with artificial intelligent agents.
I'm to tired to be coherent now; ask![[livejournal.com profile]](https://www.dreamwidth.org/img/external/lj-userinfo.gif)
talmor for proper info :-)
The gist of the talk was that (a) space is big, (b) we're going to be collecting an enormous amount of data in a relatively short time, and (c) all that data has to be stored, shifted, accessed, analysed, and accessed again. The numbers were insanely large; as the branch chair said (approximately): "We're used to dealing with zeroes and ones, but you've got a lot more zeroes than we're used to."
Since the project is partly a public relations excercise to convince some committee that we need to build this thing in Australia (rather than South Africa), a lot of effort has gone into the promotional materials. The speaker was geeky but accomplished, and there were bonus posters and glossy conference packs.
I myself am not a numbers girl. Tetra-, petra- and exo-flops don't say much more to me than "A lot". Where I can translate that into human and logistical terms, it's more bandwidth than the sea, and the same volume of data as internet traffic in a week. All this incoming into server farms so vast that even under the best conditions, failures are predicted at one per day. The practicalities of getting so much data to Perth from the Murchison before the cabling is done mean that they're seriously considering using utes once a week to collect that week's data and set up the new drives. The projections of computer grunt required run right along to the top edge of projections for the next 10 years. This with be FAST.
We were also promised 10 gague fencing wire and gaffer tape
I was impressed and saw through the vast quantity of the numbers to the infrastructure and people required to keep all this running. Yes, I'm attached to the human element in large enterprises :-)
I'm surprised they haven't started calling it SkyNet yet, and borrowing all this vast computing power to experiment with artificial intelligent agents.
I'm to tired to be coherent now; ask
talmor for proper info :-)
no subject
That's why there's going to be ~10 MegaWatt's worth of computing power to do as much analysis and reduction of that data in realtime as possible, so you don't need to store it all.
no subject