Beowulf clusters get built to support your application, not the other way around. Your choice of hardware and OS will depend on the parallel nature of your code. Do you need myrinet, or can you get away with fast ethernet? Will your code even compile under win32? Do the supporting libraries (PVM/MPI/BLAS whatever) run under win32? What about the queuing system?
How are you going to manage the cluster? You need automation, even for small clusters. How easy is it to add a new user, apply a patch or change a bios setting on your cluster without having to plug a keyboard and monitor into each node? What about central logging? How about automated OS installs when you add another 100 nodes when you get your funding?
Oh. Benchmark, benchmark, benchmark. That means your code, running your datasets, on your hardware and OS. Not vendor supplied numbers. If you have a serious hardware vendor, you should be able to wrangle demo mechines off them. Try before you buy.
The study of evolutionary relationships between genes and protein structures is a well established field called Phylogenetics.
You can get an idea of how organisms are related in evolutionary terms by comparing the DNA sequences of genes in different organisms. The classic example is cytochrome C, a protein which is vital to energy utilisation, and is found in every organism, from primates to bacteria.
If you compare the sequence of either the gene or the protein from two different spieces you will find that the sequences are not the same. Furthermore the similarity between species is not constant; chimp and human cytochrome C are almost identical, whereas human and yeast cytochrome C sequences shows a much greater number of differences. If you analyse the differences you will find that they are not random, but point to an ordered relationship between species.
The explanation for these observations is that silent mutations (ie ones which have no effect on the funtionality of the molecule) accrue at random over time. The longer the time since two species diverged, then the more random changes will occur in their gene sequences. By cross referencing the differences between many different genes for different species it is possible to build up an evolutionary tree based purely on gene sequences.(note that this is independant to the trees drawn up by comparing the fossil record, but comes to the same conclusion, a strong indication that evolutionary theory might just be true:-))
You can also try and workout a timeframe for evolution by estimating how often random mutations occur; however, this extension to the method is controvoersial, as it is not clear if mutation rates are stable across time or species.
Brittanic.com has an excellent summary, including pictures of evolutionary trees.
He was flying a U2 spy-plane over the Soviet Union and got shot down. I can't remember off hand how high the U2's flew, but it was supposed to be high enough to avoid getting hit by SAMs. Obviously he wasn't though...
After a national competition 118 russians won the chance to live inside submarine stuck at the bottom of the Barents Sea.
Then each day they must vote on who gets to the food, water and air. The losers get ejected into the sea. The particpants are under constant scrutiny from the worlds media AND 1000 feet of arctic water.
They're sole method of communication with the outside world is banging on the side of the metal hull. The winner is the one left surviving at the end and will receive $100,000,a Ford Probe and media licensing.
Today's task was to connect up an electric generator in the dark.
Apart from the ones which use R10000 and R12000, which are both 64bit processors.
Re:Differences in benchmarks can be bug symptoms
on
AMD's Duron Birthed
·
· Score: 1
Ok, it was a badly worded comment. The important thing is to know how 'small' a small error is, and whether or not it is significant.
You are always going to get some errors in numerical code, and worrying about differences in the 10th decimcal place of your calculation if your integrator is only accurate to 3 decimal places is pointless. Of course, if you are seeing larger errors than those introduced by your numerical code, *then* you have problems.
During WWII army intelligence were able to identify individual enemy radio operators from intercepted morse signals, due to the fact that each opererator had a distinctive style, known as a fist.
Given that this was possible in 1940 with no computing power, biometrics based on keyboard style is probably not so stupid...
One of the weaknesses for beowulfs seems to me to be a lack of decent (job) management software. How do you split the clusters resources? Do you run one large simulation on all the CPUs, or do you run 2 or 3 jobs on 1/2 or 1/3 of the available CPUs?
Is there provision for shifting jobs onto different nodes if one of them dies during a run?
The only good benchmark is the one when you test a specific application you are interested in running on your particular hardware when it sits on your network running under your normal workload.
The EBI has a list of the various genome databases, which are all freely searchable. You can check up on the progress of the various genome sequencing projects here.
How can we pretend to have an informed debate when the standard of knowledge is so poor? Katz asks why there has been little debate over Venter's discovery of a minimal genome. Probably because there isn't really anything to debate. What is so special about creating life? As far as I remember, the debate over 'vitalism' (that living matter is somehow different to the inanimate suff) was sorted out a little over 100 years ago. Katz might want to update his ideas a little bit. The ability to create a novel organism doesn't grant anyone any mystical status other than to get their name on a few scientific papers.
Furthermore, plant and animal breeders have been creating "new species" in a rather more piece-meal and haphazard fashion for the past few thousand years. Oh, and clones (aka idential twins, and at a pinch, any creature which happens to reproduce asexually) are all around us. Does this spell doom and destruction for all of us?
All that hysteria does is prevent rational discourse about the important issues. There are privicy issues that are going to arise about the use of genetic details obtained about individuals. However, many countries (esp in the EU) already have a legal framework in place to deal with the disemination of personal data collected by companies to third parties.
The completion of the human genome project is not going to spell the end of civilsation as we know it. Firstly, a genetic sequence is just that, a string of letters. Actually finding out what a gene does, let alone working out how to modulate it function is not a trivial task.
Another question you might like to ask yourself is why is genetic manipulation seen as an automatically bad thing? Is having gene therapy to control chronic obesity worse than getting liposuction? Or altering the genes that control nose shape worse than having a good, old-fashioned nose-job?
What is needed is a lot more thought when someone mentions genetic-engineering, and a lot less hand-wringing.
Re:Well, this is the kind of thing you have to exp
on
Caught Before the Act
·
· Score: 1
Gun control was not introduced by "Due democratic" process. It was mob rule imposed by vote chasing, populist MPs. Popular laws are not neccessarily just laws.
Banning hand-guns has had negligible effect on gun crime in the UK. Nobody robs banks or takes part in a gangland turf war with a legally registered handgun.
Our SGIs are named after the fireman from Trumpton; Pugh, Pugh, (its a multiprocessor machine),Barney, Mcgrew, Cuthbert, Dibble and Grub. Our Decs are named after indian foods, so we have Balti and Bhaji.
The Meterological Office used to have a pair of Crays called Ronnie and Reggie...
I've always found VRML to be the 3D equivalent of HTML; Give somone a document in HTML and you know they are going to be able to view it, no matter what OS they are using. Same thing for VRML if you need to share 3D objects. As was mentioned above, it is especially useful for scientific visualisation, especially when you can never be sure that your collaborators are using the same modelling/visualisation package as you are. Think of it as a lowest common denomenator.
(shameless plug): I've got some VRML molecules on my web page http://www.biochemistry.bham.ac.uk/gcoates/resea rch.html
If you think a standard for tea is bad, a quick search on http://bsonline.techindex.co.uk reveals BS0-1:1997, "A British Standard for Standards." I wonder if BS0-1:1997 is BSO-1:1997 compliant?
The Dangerous Sports Club took one under Tower Bridge in London back in the 80s. Apparently wind resistance on those things is considerable, and a gust of wind made the ball break free from the tugs that were trying to tow it. The people inside almost drowned as the thing slowly collapsed and filled with nice, clean Thames water. Still, a miss is as good as a mile...
If you want a to know about free will, you have to understand the quantum nature of time. David Deutsch's excellent book, "The fabric of reality" does a very good job of explaining quantum time, and its concequences for parallel universes, free will and quantum computation.
Alphas are the dog's dangly bits if you are doing any sort of numerically intensive work. However, performance is typically better under OSF1/Tru64/Digital Unix than linux, as the DEC native compiler does a better job of optimising code, and has a faster maths library than gcc.
Quite a few labs tend to fork out for third party compilers and librarys for linux, in order to get the best out of the hardware. I hear it is possible to compile static binaries under OSF1 and run them under alpha-linux though...
In order to keep traffic moving on the M25 (London orbital motorway) they have introduced variable speed limits. As the traffic get heavier, the speed limit gets lower. Supposedly this stops the traffic flow from becoming "turbulent", and so everything keep moving.
Unfortunately, the way it is enforced is rather scary. They have electronic cameras placed at intervals along the road. The cameras do OCR on each car's numberplate as it passes, and if your average speed exceeds the limit, you get a fine+points through the post a few days later. Being digital, the cameras never run out of film, and as they track your average speed, the normal trick of slowing down when you see a camera and speeding up inbetween does not work.
Apparantly the caught 4300 speeders on the first day the system went live. And of course, tracking cars all over London via their numberplate isn't an infringment of civil liberties. Honest.
Slashdot Mirror
IAABA. (I am a beowulf admin).
Beowulf clusters get built to support your application, not the other way around. Your choice of hardware and OS will depend on the parallel nature of your code. Do you need myrinet, or can you get away with fast ethernet? Will your code even compile under win32? Do the supporting libraries (PVM/MPI/BLAS whatever) run under win32? What about the queuing system?
How are you going to manage the cluster? You need automation, even for small clusters. How easy is it to add a new user, apply a patch or change a bios setting on your cluster without having to plug a keyboard and monitor into each node? What about central logging? How about automated OS installs when you add another 100 nodes when you get your funding?
Oh. Benchmark, benchmark, benchmark. That means your code, running your datasets, on your hardware and OS. Not vendor supplied numbers. If you have a serious hardware vendor, you should be able to wrangle demo mechines off them. Try before you buy.
The study of evolutionary relationships between genes and protein structures is a well established field called Phylogenetics.
:-))
You can get an idea of how organisms are related in evolutionary terms by comparing the DNA sequences of genes in different organisms. The classic example is cytochrome C, a protein which is vital to energy utilisation, and is found in every organism, from primates to bacteria.
If you compare the sequence of either the gene or the protein from two different spieces you will find that the sequences are not the same. Furthermore the similarity between species is not constant; chimp and human cytochrome C are almost identical, whereas human and yeast cytochrome C sequences shows a much greater number of differences. If you analyse the differences you will find that they are not random, but point to an ordered relationship between species.
The explanation for these observations is that silent mutations (ie ones which have no effect on the funtionality of the molecule) accrue at random over time. The longer the time since two species diverged, then the more random changes will occur in their gene sequences. By cross referencing the differences between many different genes for different species it is possible to build up an evolutionary tree based purely on gene sequences.(note that this is independant to the trees drawn up by comparing the fossil record, but comes to the same conclusion, a strong indication that evolutionary theory might just be true
You can also try and workout a timeframe for evolution by estimating how often random mutations occur; however, this extension to the method is controvoersial, as it is not clear if mutation rates are stable across time or species.
Brittanic.com has an excellent summary, including pictures of evolutionary trees.
He was flying a U2 spy-plane over the Soviet Union and got shot down. I can't remember off hand how high the U2's flew, but it was supposed to be high enough to avoid getting hit by SAMs. Obviously he wasn't though...
Does anyone know how high Gary Powers was when he bailed out of his U2? That must have been a pretty high jump.
Meanwhile in Russia
After a national competition 118 russians won the chance to live inside submarine stuck at the bottom of the Barents Sea.
Then each day they must vote on who gets to the food, water and air. The losers get ejected into the sea. The particpants are under constant scrutiny from the worlds media AND 1000 feet of arctic water.
They're sole method of communication with the outside world is banging on the side of the metal hull. The winner is the one left surviving at the end and will receive $100,000,a Ford Probe and media licensing.
Today's task was to connect up an electric generator in the dark.
Area is measured in wembley stadiums, which also doubles as a measure of volume. However, I think pints of beer is the official SI volume unit.
Might be a good thing if they mean to migrate people to use ssh instead. Getting people weaned off telnet is hard unless:
a) A machine gets cracked and their data get hosed.
b) You disable telnet and force them to use ssh.
Of course, if they mean 'We should not allow people to access remote computers at any time' then they need hitting with the cluestick.
>The SGI O2 uses some MIPS R5000 CPU.
Apart from the ones which use R10000 and R12000, which are both 64bit processors.
Ok, it was a badly worded comment. The important thing is to know how 'small' a small error is, and whether or not it is significant.
You are always going to get some errors in numerical code, and worrying about differences in the 10th decimcal place of your calculation if your integrator is only accurate to 3 decimal places is pointless. Of course, if you are seeing larger errors than those introduced by your numerical code, *then* you have problems.
You should expect small differences in the results of numerical calcs when running the same code on different OS/hardware/compiler combinations.
Presumably the accuracy of the numerical integrator is going to be greater than the differences in the FP precision?
During WWII army intelligence were able to identify individual enemy radio operators from intercepted morse signals, due to the fact that each opererator had a distinctive style, known as a fist.
Given that this was possible in 1940 with no computing power, biometrics based on keyboard style is probably not so stupid...
One of the weaknesses for beowulfs seems to me to be a lack of decent (job) management software. How do you split the clusters resources? Do you run one large simulation on all the CPUs, or do you run 2 or 3 jobs on 1/2 or 1/3 of the available CPUs?
Is there provision for shifting jobs onto different nodes if one of them dies during a run?
Preliminary linux drivers can be found here:
http://hem.fyristorg.com/henrikj/em8300/
The only good benchmark is the one when you test a specific application you are interested in running on your particular hardware when it sits on your network running under your normal workload.
Nuff said.
The EBI has a list of the various genome databases, which are all freely searchable. You can check up on the progress of the various genome sequencing projects here.
The DRI readme states that DRI will only be
for i386 at the moment. Us lucky alpha owners
will have to wait to get decent 3D acceleration.
Furthermore, plant and animal breeders have been creating "new species" in a rather more piece-meal and haphazard fashion for the past few thousand years. Oh, and clones (aka idential twins, and at a pinch, any creature which happens to reproduce asexually) are all around us. Does this spell doom and destruction for all of us?
All that hysteria does is prevent rational discourse about the important issues. There are privicy issues that are going to arise about the use of genetic details obtained about individuals. However, many countries (esp in the EU) already have a legal framework in place to deal with the disemination of personal data collected by companies to third parties.
The completion of the human genome project is not going to spell the end of civilsation as we know it. Firstly, a genetic sequence is just that, a string of letters. Actually finding out what a gene does, let alone working out how to modulate it function is not a trivial task.
Another question you might like to ask yourself is why is genetic manipulation seen as an automatically bad thing? Is having gene therapy to control chronic obesity worse than getting liposuction? Or altering the genes that control nose shape worse than having a good, old-fashioned nose-job?
What is needed is a lot more thought when someone mentions genetic-engineering, and a lot less hand-wringing.
Gun control was not introduced by "Due democratic" process. It was mob rule imposed by vote chasing, populist MPs. Popular laws are not neccessarily just laws.
Banning hand-guns has had negligible effect on gun crime in the UK. Nobody robs banks or takes part in a gangland turf war with a legally registered handgun.
Our SGIs are named after the fireman from Trumpton;
Pugh, Pugh, (its a multiprocessor machine),Barney, Mcgrew, Cuthbert, Dibble and Grub.
Our Decs are named after indian foods, so we have Balti and Bhaji.
The Meterological Office used to have a pair of Crays called Ronnie and Reggie...
I've always found VRML to be the 3D equivalent of HTML; Give somone a document in HTML and you know they are going to be able to view it, no matter what OS they are using. Same thing for VRML if you need to share 3D objects. As was mentioned above, it is especially useful for scientific visualisation, especially when you can never be sure that your collaborators are using the same modelling/visualisation package as you are. Think of it as a lowest common denomenator.
a rch.html
(shameless plug): I've got some VRML molecules on my web page
http://www.biochemistry.bham.ac.uk/gcoates/rese
If you think a standard for tea is bad, a quick search on http://bsonline.techindex.co.uk reveals BS0-1:1997,
"A British Standard for Standards."
I wonder if BS0-1:1997 is BSO-1:1997 compliant?
The Dangerous Sports Club took one under Tower Bridge in London back in the 80s. Apparently wind resistance on those things is considerable, and a gust of wind made the ball break free from the tugs that were trying to tow it. The people inside almost drowned as the thing slowly collapsed and filled with nice, clean Thames water.
Still, a miss is as good as a mile...
If you want a to know about free will, you have to
understand the quantum nature of time. David Deutsch's excellent book, "The fabric of reality" does a very good job of explaining quantum time, and its concequences for parallel universes, free will and quantum computation.
Alphas are the dog's dangly bits if you are doing any sort of numerically intensive work. However, performance is typically better under OSF1/Tru64/Digital Unix than linux, as the DEC native compiler does a better job of optimising code, and has a faster maths library than gcc.
Quite a few labs tend to fork out for third party compilers and librarys for linux, in order to get the best out of the hardware. I hear it is possible to compile static binaries under OSF1 and run them under alpha-linux though...
In order to keep traffic moving on the M25 (London orbital motorway) they have introduced variable speed limits. As the traffic get heavier, the speed limit gets lower. Supposedly this stops the traffic flow from becoming "turbulent", and so everything keep moving.
Unfortunately, the way it is enforced is rather scary. They have electronic cameras placed at intervals along the road. The cameras do OCR on each car's numberplate as it passes, and if your average speed exceeds the limit, you get a fine+points through the post a few days later. Being digital, the cameras never run out of film, and as they track your average speed, the normal trick of slowing down when you see a camera and speeding up inbetween does not work.
Apparantly the caught 4300 speeders on the first day the system went live. And of course, tracking cars all over London via their numberplate isn't an infringment of civil liberties. Honest.