A little question, No java expert so how can I do this, can´t find a file to start topcat with.
Hi!
But you downloaded the *.jar file and you have java 1.5 or better installed? Then
java -jar {name-of-jar-file}
will start an application that's packaged as a single jar.
Have fun
H-BE
Thanks now I got it to work, somehow when I tried to download the *.jar file my computer automaticaly renamed it to *.zip, but when I renamed to *.jar and double clicked on it everything work perfectly.
Visualizing the results is fun, but the result files get uploaded really fast and I really don't feel comfortable snooping around in the slots directories while the WUs are still processed.
So I wondered whether it might not be a better approach to collect the results in one central place for my whole LAN: technically it would be a HTTP proxy server that would copy intercept the file uploads, forward them to E@H (of course) but would also unpack the data and store it somewhere (database, files, whatever).
I wonder whether there are already BOINC related open source proxy servers, e.g. for collecting statistics or for detecting and diagnosing problems in bigger BOINC networks.
EDIT: ...provided the uploads are done via HTTP at all. It's not done via ftp, is it?
...
EDIT: ...provided the uploads are done via HTTP at all. It's not done via ftp, is it?
CU
Bikeman
as far as I know, only port 80 is used to communicate with the BOINC servers...
(firewall connection through port 21/23 would not be allowed)
Yup, I've finally managed to eavesdrop on an upload (more fun than looking it up in the source code :-). It's really done thru a standard HTTP POST request, fine with me.
I found a rather lightweight open source HTTP proxy implemented in JAVA (called Muffin) which is easy to extend and well written, so I guess all the results generated within my home network will be dumped in a central repository soon.
BTW, not all results look as "boring" as the example form S5R3 in the first post, which appears to be just a ring of uniformly distributed dots along a latitude circle.
Some look rather clustered around one or two points in the sky, like this one:
(the red stuff is an overlay of known pulsars similar to those shown in the screensaver view, for orientation). What you see here is that almost all of the 10000 sky points returned to the server are located in one tight cluster (the rainbow-colored stuff in the lower left quadrant).
Well, if you see something like this, don't immediately call the New York Times or wake up Bruce Allen in the middle of the night claiming to have found a gravitational wave :-).
See Bernd's message above stating that the result that gets sent back is just the raw input for the next stage in a hierarchical search scheme. Only after that second stage, you can tell what is a real signal and what isn't. Anyway, what we see here is (I guess) the same effect that is described here: in certain frequency bands, oscillations in the suspension of mirrors (and other "noise") in the detectors introduce "false candidate" pulsar signals. Still, it's cool to see that not all results are equal in appearance.
The higher the frequency band, the finer we need to make the grid of templates. As by design each workunit has basically the same number of templates, in higher frequency bands there aren't enough to span a whole circle, so it's no wonder the most significant candidates of some workunits are limited to a certain area.
Whether such a spot actually is significant can only be told if taking the results of the adjacent workunits into comparison.
The higher the frequency band, the finer we need to make the grid of templates. As by design each workunit has basically the same number of templates, in higher frequency bands there aren't enough to span a whole circle, so it's no wonder the most significant candidates of some workunits are limited to a certain area.
This particular result was for 418.95Hz with still enough skypoints per WU to form a closed ring near the poles (it's a bit blurred in the image but if you look real close you can see a few blue dots on the opposite side of the "cluster".
However,
Quote:
Whether such a spot actually is significant can only be told if taking the results of the adjacent workunits into comparison.
BM
so still no reason to call the New York Times, I'm afraid :-)
so still no reason to call the New York Times, I'm afraid :-)
No, not at all. The are other possible reasons for sucha spot, even if it should come out statistically significant: It can be an instrumental artifact (harmonics of 60Hz are pretty noisy, for obvious reasons), or a fake signal we injected into the data to validate the analysis (yes we do).
This is a set of results for the 343.55 Hz frequency band.
But again, this is a fake signal. The signal appears at a sky position where a real pulsar signal would produce the weakest Doppler drift, and the frequency is from a band of frequencies that is known to produce false candidates, see here, because some suspension wires of the optical instruments in the LIGO interferometers tend to oscillate at those frequencies.
RE: RE: RE: Hi! 7)
)
Thanks now I got it to work, somehow when I tried to download the *.jar file my computer automaticaly renamed it to *.zip, but when I renamed to *.jar and double clicked on it everything work perfectly.
Thanks
Hi! Visualizing the
)
Hi!
Visualizing the results is fun, but the result files get uploaded really fast and I really don't feel comfortable snooping around in the slots directories while the WUs are still processed.
So I wondered whether it might not be a better approach to collect the results in one central place for my whole LAN: technically it would be a HTTP proxy server that would copy intercept the file uploads, forward them to E@H (of course) but would also unpack the data and store it somewhere (database, files, whatever).
I wonder whether there are already BOINC related open source proxy servers, e.g. for collecting statistics or for detecting and diagnosing problems in bigger BOINC networks.
EDIT: ...provided the uploads are done via HTTP at all. It's not done via ftp, is it?
CU
Bikeman
RE: ... EDIT: ...provided
)
as far as I know, only port 80 is used to communicate with the BOINC servers...
(firewall connection through port 21/23 would not be allowed)
Udo
RE: RE: ... EDIT:
)
Yup, I've finally managed to eavesdrop on an upload (more fun than looking it up in the source code :-). It's really done thru a standard HTTP POST request, fine with me.
I found a rather lightweight open source HTTP proxy implemented in JAVA (called Muffin) which is easy to extend and well written, so I guess all the results generated within my home network will be dumped in a central repository soon.
CU
Bikeman
Hi! BTW, not all results
)
Hi!
BTW, not all results look as "boring" as the example form S5R3 in the first post, which appears to be just a ring of uniformly distributed dots along a latitude circle.
Some look rather clustered around one or two points in the sky, like this one:
(the red stuff is an overlay of known pulsars similar to those shown in the screensaver view, for orientation). What you see here is that almost all of the 10000 sky points returned to the server are located in one tight cluster (the rainbow-colored stuff in the lower left quadrant).
Well, if you see something like this, don't immediately call the New York Times or wake up Bruce Allen in the middle of the night claiming to have found a gravitational wave :-).
See Bernd's message above stating that the result that gets sent back is just the raw input for the next stage in a hierarchical search scheme. Only after that second stage, you can tell what is a real signal and what isn't. Anyway, what we see here is (I guess) the same effect that is described here: in certain frequency bands, oscillations in the suspension of mirrors (and other "noise") in the detectors introduce "false candidate" pulsar signals. Still, it's cool to see that not all results are equal in appearance.
CU
Bikeman
The higher the frequency
)
The higher the frequency band, the finer we need to make the grid of templates. As by design each workunit has basically the same number of templates, in higher frequency bands there aren't enough to span a whole circle, so it's no wonder the most significant candidates of some workunits are limited to a certain area.
Whether such a spot actually is significant can only be told if taking the results of the adjacent workunits into comparison.
BM
BM
RE: The higher the
)
This particular result was for 418.95Hz with still enough skypoints per WU to form a closed ring near the poles (it's a bit blurred in the image but if you look real close you can see a few blue dots on the opposite side of the "cluster".
However,
so still no reason to call the New York Times, I'm afraid :-)
CU
H-B
RE: so still no reason to
)
No, not at all. The are other possible reasons for sucha spot, even if it should come out statistically significant: It can be an instrumental artifact (harmonics of 60Hz are pretty noisy, for obvious reasons), or a fake signal we injected into the data to validate the analysis (yes we do).
BM
BM
RE: (harmonics of 60Hz are
)
Amazing, I completed a result today in the same sky region but 60 Hz up ... guess what ;-)
I guess this alone is a good reason for having GW observatories in the US and Europe, where 50Hz is the usual AC frequency.
CU
H-B
Another dramatic looking
)
Another dramatic looking result:
This is a set of results for the 343.55 Hz frequency band.
But again, this is a fake signal. The signal appears at a sky position where a real pulsar signal would produce the weakest Doppler drift, and the frequency is from a band of frequencies that is known to produce false candidates, see here, because some suspension wires of the optical instruments in the LIGO interferometers tend to oscillate at those frequencies.
Bikeman