I was interviewed for a PhD place (which I didn't get) in 1994 which would have been analysing "bottom simulating reflectors" in seismic reflection data from the Vancouver Island area. It was already known that these "reflectors" represented methyl hydrate deposits, and this was one of the reasons put forward on the grant applications.
Although the PhD really concerned geophysical techniques, the background methyl-hydrate argument supporting the project was actually concerned about global warming (warming seas -> destabilisation of methyl hydrate => might be a good idea to know how much and how stable the stuff is) rather than a future fuel source.
And for active seismic data, 40MB is pretty small.
For the type of thing you're talking about, it is after a lot of processing.
The actual processing of the seismic data (to get the interpretation data) divides into three categories:
- bulk "simple". Calculations are simple, essentially things like addition, scaling, etc
Large quantities of data - Raw field tapes - maybe a few TB/dataset these days. Bottleneck is typically the disks and/or the tapes.
(in 1998 where I worked, they had over a petabyte of tape store in the machine room - that excludes the data archive!)
- "big iron". Bottleneck are the cpus & main bus. Only a few processing tasks come into this area, but it is well worth buying a supercomputer to do it. Luckily such tasks are typically very parallel, and often suitable for a vector processor. Of course, using such techniques means you need fast buses/etc which is where the above announcement comes in.
- "interactive". Typically, a small section of the dataset is selected from tape to a fast disk (or even RAM). Processing is on the fly, as the user chooses the best parameters.
Often these parameters are then used for batch processing one of the previous categories.
RB
Re:kinda takes the PUBLIC out of it doesn't it?
on
Blogspace vs. NPR
·
· Score: 1
They also tried to campaign against low-power community radio stations.
One of the claimed advantages of such stations is that they're very local and owned/operated by the populace - rather than some national company.
RB
Slashdot Mirror
I was interviewed for a PhD place (which I didn't get) in 1994 which would have been analysing "bottom simulating reflectors" in seismic reflection data from the Vancouver Island area. It was already known that these "reflectors" represented methyl hydrate deposits, and this was one of the reasons put forward on the grant applications.
Although the PhD really concerned geophysical techniques, the background methyl-hydrate argument supporting the project was actually concerned about global warming (warming seas -> destabilisation of methyl hydrate => might be a good idea to know how much and how stable the stuff is) rather than a future fuel source.
RB
The actual processing of the seismic data (to get the interpretation data) divides into three categories:
- bulk "simple". Calculations are simple, essentially things like addition, scaling, etc Large quantities of data - Raw field tapes - maybe a few TB/dataset these days. Bottleneck is typically the disks and/or the tapes. (in 1998 where I worked, they had over a petabyte of tape store in the machine room - that excludes the data archive!)
- "big iron". Bottleneck are the cpus & main bus. Only a few processing tasks come into this area, but it is well worth buying a supercomputer to do it. Luckily such tasks are typically very parallel, and often suitable for a vector processor. Of course, using such techniques means you need fast buses/etc which is where the above announcement comes in.
- "interactive". Typically, a small section of the dataset is selected from tape to a fast disk (or even RAM). Processing is on the fly, as the user chooses the best parameters. Often these parameters are then used for batch processing one of the previous categories.
RB
They also tried to campaign against low-power community radio stations. One of the claimed advantages of such stations is that they're very local and owned/operated by the populace - rather than some national company. RB