What amount of ram? Like the amount you'd need to add together 196 numbers?
Aligning the images will probably happen in pairs... Once you have all the offsets computed between each image and it's anchor image (say anchor is to the left, and for the leftmost column it's to the top), then you can just fly through all the images reading a row (or two) at a time and spitting out the result.
You'll have to load all the images to do the alignment, and it may work better if you used a wider alignment (image plus all 8 around it), but still that's only 9 images loaded at once, followed by the next 9 images.
We're in a similar situationat a large public university I work at. There's been talk of trying to get some sort of site-license deal setup for RHN. From what I've heard it looks VERY reasonable.
You act as if this is new news. It isn't. This has been known for a while -- at least since august, because that's when my group first started discussing what we were going to do about it.
The really great thing is we just had a Microsoft security speaker at the ACM Reflections|Projections conference at UIUC.
He was talking about how important it is to have secure code, and all the initiatives they have to fix security holes.
He also talked about how fast worms are spreading these days. Patching is not going to be sufficient - a bug discovered and posted will turn into a worm hours or days before Microsoft will respond with a patch. By then it'll be too late.
The speed of the interconnect puts a limit on the number of nodes you can practically connect.
Given our CPU power has been growing far faster than the networking speeds, he chose well in that aspect. Low-latency communication is vital to most (not all) parallel applications.
That depends how much a maintinence nightmare it ends up being. Time will tell. Raw number crunching is great but is not the end all be all of clusters, as strange as that may sound.
(UIUC) Notes is in fact still the only tolerable message board in existance. At least compared to all that web crap today. Nobody who likes unix could say they like the web crap more than good 'ol command line notes.
Perdicting the weather of a small or large area makes no difference. Weather simulation accuracy improves as the side of the grid you use decreases. It doesn't matter if you're predicting for the entire USA or just one state, if you don't get detailed enough you're off period.
Okay so you have your data on the remote machine encrypted with your PGP key. Which you kept on your local machine, or maybe you kept it on a floppy or usb keyring by your local machine.
Disaster strikes. Byebye local machien, byebye PGP key. How exactly do you recover now?
Just like they do now, duh. The article even tells you how they'll do it if you think about it.
The numbers they publish right now are Mhz or Ghz or whatever. Those numbers are a measure of how fast the slowest element in the chip can do it's work. (well 1/how fast...)
The article said the async chips it doesn't matter about the slowest, but about the average. So they compute the average (or an approximination of such) time taken to do a unit of work, divide 1 by than and there ya go, instant Mhz-equiv like AMD is doing right now with the athlon.
Reguardless, the Jag has 2 games that make it all worthwhile to own a system:
Tempest2k, and Rayman.
At the time they came out, Rayman didn't exist for the PC (might have been on other consoles), and tempest sure didn't. Two of the better games out there, IMO.
We can't power the US's demand on corn energy. Hell even a state like IL could power I believe roughly 10% (calculated in an energy class I took for the hell of it. Actually learned stuff too, was impressed. And a tour of a nuclear reactor is always cool. Not like that's why i sugned up for it. Of course not!) of it's petrolium based energy needs via ethanol if all the farmland in IL was used for nothing but ethanol production.
Slashdot Mirror
What amount of ram? Like the amount you'd need to add together 196 numbers?
Aligning the images will probably happen in pairs... Once you have all the offsets computed between each image and it's anchor image (say anchor is to the left, and for the leftmost column it's to the top), then you can just fly through all the images reading a row (or two) at a time and spitting out the result.
You'll have to load all the images to do the alignment, and it may work better if you used a wider alignment (image plus all 8 around it), but still that's only 9 images loaded at once, followed by the next 9 images.
So exactly how many processors was your database ment to run on? 128? More? Did it expect each CPU to have more than about 32-64M of memory?
BG/L is looking like a great supercomputer, but it probably would make a truly sucky anything-else-server.
We're in a similar situationat a large public university I work at. There's been talk of trying to get some sort of site-license deal setup for RHN. From what I've heard it looks VERY reasonable.
You act as if this is new news. It isn't. This has been known for a while -- at least since august, because that's when my group first started discussing what we were going to do about it.
You could always, you know, roll your own RPMs.
The really great thing is we just had a Microsoft security speaker at the ACM Reflections|Projections conference at UIUC.
He was talking about how important it is to have secure code, and all the initiatives they have to fix security holes.
He also talked about how fast worms are spreading these days. Patching is not going to be sufficient - a bug discovered and posted will turn into a worm hours or days before Microsoft will respond with a patch. By then it'll be too late.
It's called government grant money.
The speed of the interconnect puts a limit on the number of nodes you can practically connect.
Given our CPU power has been growing far faster than the networking speeds, he chose well in that aspect. Low-latency communication is vital to most (not all) parallel applications.
For the sake of humanity, we can only hope.
That depends how much a maintinence nightmare it ends up being. Time will tell. Raw number crunching is great but is not the end all be all of clusters, as strange as that may sound.
RDS was in fact very cool.
Now all we need is someone to re-release something like M.U.L.E. and I'll be happy.
Which'd be insightful if it were the hardware they are guarding, not the data on it.
(UIUC) Notes is in fact still the only tolerable message board in existance. At least compared to all that web crap today. Nobody who likes unix could say they like the web crap more than good 'ol command line notes.
SVIRP RID ON!
Ignoring the unix updating tools, it's hard to update.
No kidding. Let me guess, ignoring the sun, it's dark?
Uh, no? A turing machine has infinite memory.
You have to be a slow typist to need more than 20 seconds to make a correction to two lines of text.
Perdicting the weather of a small or large area makes no difference. Weather simulation accuracy improves as the side of the grid you use decreases. It doesn't matter if you're predicting for the entire USA or just one state, if you don't get detailed enough you're off period.
Okay so you have your data on the remote machine encrypted with your PGP key. Which you kept on your local machine, or maybe you kept it on a floppy or usb keyring by your local machine.
Disaster strikes. Byebye local machien, byebye PGP key. How exactly do you recover now?
Then run more conservative programs to go slower. In the end you go the same speed as everyone else but spend tons of useless cooling solutions.
Hmm.
Seems to work around it fine.
And don't let the door hit you on the way out.
Otherwise we'd hve to call it GNU/BeOS.
Tho I may have mispelled it. It's a mini-cave system, complete with tracker and small enough to fold into a crate for shipping.
CAVEs can also be shipped. Setup and teardown time is pretty harsh but it can (and has at NCSA/UIUC) been done.
Nonono you have it all wrong.
Pepsi blue is SMURF!
Just like they do now, duh. The article even tells you how they'll do it if you think about it.
The numbers they publish right now are Mhz or Ghz or whatever. Those numbers are a measure of how fast the slowest element in the chip can do it's work. (well 1/how fast...)
The article said the async chips it doesn't matter about the slowest, but about the average. So they compute the average (or an approximination of such) time taken to do a unit of work, divide 1 by than and there ya go, instant Mhz-equiv like AMD is doing right now with the athlon.
What part of the equation isn't clear to you?
Reguardless, the Jag has 2 games that make it all worthwhile to own a system:
Tempest2k, and Rayman.
At the time they came out, Rayman didn't exist for the PC (might have been on other consoles), and tempest sure didn't. Two of the better games out there, IMO.
We can't power the US's demand on corn energy. Hell even a state like IL could power I believe roughly 10% (calculated in an energy class I took for the hell of it. Actually learned stuff too, was impressed. And a tour of a nuclear reactor is always cool. Not like that's why i sugned up for it. Of course not!) of it's petrolium based energy needs via ethanol if all the farmland in IL was used for nothing but ethanol production.