Right, they left out SGI, which happens to be a significant player in the gov field. Why is it that SGI is bypassed when people talk commercial Unix? This still mystifies me.
Worse, the evaluation placed high emphasis on admin tools. Around here, with Solaris, Irix, HP-UX, Digital Unix and Linux in one shop, noone is ever going to use SMIT or SAM or whatever. Just too much pain to learn them all.
Unifying sysadmin tasks is best done with scripting, at which Unix excels (as opposed to NT, say).
If your shop has one flavor of Unix only and has few, if any experienced sysadmins, then admin tools are for you. However, I don't think that most shops in the gov domain are small and hence emphasis on administration ease seem misguided.
What equally mystifies me is that the Compaq unit beats the competition 2:1 on CPU benchmarks, but only gets 3 stars on performance. Yes, it was weak in the graphics dept., but how can you loose so badly that you give up a 2:1 advantage? In my workstation I'd rather tolerate weakish graphics than a slow CPU. Of course, we are never told actual performance numbers and how they were weighted. Too bad.
The article quotes Alan Baratz of SUN, not someone inside AOL/Netscape. AFAIK, SUN does not have its hands significantly in the browser pie, and Baratz is not at the helm of the combined SUN/AOL effort either. So, that piece of journalism is just FUD. Besides, the way I read the netscape/mozilla license, the code can't be closed anymore.
Its funny how SUN people push their misguided Community license stuff. It just won't fly. It's just a cheap attempt at grabbing attention and get a few gullible souls to work for them for free.
Re:Some thoughts/comparisons
on
1GHz Alphas
·
· Score: 1
We have had an opportunity to benchmark HP's new N4000 server with 8 440 MHz PA-8500 CPUs in them. The results in a nutshell: The PA-8500 inside that server lives up to what its SPECfp promises. It is 1.5-2x faster than SGI Origin2000 with 250 MHz MIPS R10000. That's with computational chemistry / molecular modelling codes crunching real world data sets up to 4 way parallel.
So, in contrast to what the above poster guessed, SPECfp does not seem to be skewed when compared to real world performance. I'd like to see how the machine scales up to 8way. Soon we'll know that too.
This is nothing new really. It is exactly what our immune system is doing already, and very well at that. The immune system even continues to function when being hammered with myriads of problems all at the same time or when receiving a strong do not act signal. And the immune system is just a part of the continuous growth and turnover in our bodies which is usually referred to as life.
Without the continous healing, rebuilding and growth occurring in our bodies, you and I would cease to exist within days, if not quicker.
Our bodies are not static. They are the product of continuous buildup and removal of molecules, precisely orchestrated in space and time and self-organizing. As such, we resemble much more the seemingly static patterns seen in flowing water than a machine that is first built and then used. See nonlinear thermodynamics and chaos theory.
I just wonder if purposely built 'machines/computers' to 'enhance' these functions will really do so, or rather get in the way.
Keep in mind that the Ultra5 is the frontend of the E10000, e.g. the console from which to manage the E10000. It has not much work to do, just needs to be here and monitor some activity. It is NOT used as a server.
Being a scientist myself, I feel that E-biomed is a great idea. Finally someone of the establishment challenges age-old ideas and tries to explore new territory. E-publishing (not the kind of 'yeah, we also offer paid access to online versions' of the journal publishers) will curb the rising costs libraries and individual subscribers face, it will do away with the signing away of copyright to publishers AND it will significantly challenge the peer review system that we are all so fond of (or are we?). Let me tackle each one:
Cost: Libraries have complained for many years about the rising costs of print subscriptions. Many have been forced to reduce the number of journals they subscribe to, all the while the number of journals being published has actually risen. Reasons for rising costs are many, but one chief contributor is the pressure on scientists to publish more - publish or perish is the key word. As a consequence, research results are being broken down into 'least publishable units' and pushed out into papers. E-publishing in the E-biomed sense, while not necessarily curbing the proliferation of published papers, will at least significantly cut the cost of publishing and accessing papers.
There is also a growing amount of data that cannot be easily published in paper form, nor does it fit established scientific databases such as Genbank. E-publishing lets us put this stuff out there and make accessible to anybody who wants it.
Copyyright: It has long irked me and many scientists I know that many publishers require me to sign a copyright transfer agreement which robs me of most rights to my own output. If I decide to reuse a figure in an overview article, I actually have to go and ask permission to do so. Granted, I ususally do get that permission, but it bugs me to no end that I have to jump through such hoops just so that I can get into a first tier journal. It was my creativity, planning, execution and insight that enabled the publication in the first place. The publisher's contribution, while not negligible, is relativeley minor. E-publishing, if done right, will let me retain copyright of all my works.
Peer review: As many other contributors to this discussion have pointed out, peer review guarantees quality of content. However, science being an endeavor of humans, peer review is fraught with perils. It is mostly an old-boy network and hence new people, new ideas, unorthodox approaches and new interpretations of old data have a very hard time getting past the reviewers. By its very nature peer review resists change. Furthermore, because 'peers' are also mostly competitors, and because the review process is confidential, the system invites fraud on the part of the reviewers. Be that stealing ideas or simply slamming a paper in order to slow down the authors, it is not in the spirit of furthering science, but definitely occurs quite often.
Now contrast peer review with the experience we have been able to gather with the pysics eprint archive. Eprints are not reviewed formally, but despite this the eprint archive is highly regarded in the field and is probably the most widely used dissemination vehicle. Why? The wide readership and instant feedback that is enabled by the system essentially guarantees that authors are very careful to not make fools of themselves. This is not an oligarchy of the establishment as in peer review, but rather democracy in action. I'll take that anyday over peer review.
Now what's really missing is free access to data collected by the state about us. Europeans have that, and it is scary how much BS needs to be corrected sometimes when an individual takes the time and effort to check. Then again, I prefer for them to sell BS to companies rather than the truth about me;-)
Not necessarily a bad thing. After all, they must have my written consent. Also think about how many times you have handed over income info voluntarily to commercial entities already.
Having said all that, I still have nagging doubts that there will be slip-ups and just plain against-the-rules actions on the part of the responsible people.
Journalling filesystems slower than non-journalling ones? Not XFS!
XFS is actually faster than its non-journalling predecessor, EFS. I had both on the same hardware (low-end Indy). EFS itself wasn't bad. I thought is was faster than SUN's ufs on otherwise comparable hardware with the apps I used back then.
I'd never go back to an old-style FS if I had a choice. Glad to see a key piece of technology open sourced.
Correct, you have to use MPI or something similar to run a single app across a cluster. BUT: There are tons of applications in science where one essentially runs 100s or 1000s of smallish calculations while varying some input parameters or input data. For these, a Linux cluster is the perfect match.
We are currently planning a 1000 CPU cluster for such an application where not even a high-speed network is needed, just lots of cycles to get the job done before we grow old.
right on! Linux needs a fair bit of work until it can run on 64 CPUs or more, handle very big files, directories and databases and such interestingly, SGI has stated that they want to migrate such features from Unicos and Irix into Linux and put the result back into the open
it'll be very interesting to watch events over the next two years or so...
Slashdot Mirror
Right, they left out SGI, which happens to be a significant player in the gov field. Why is it that SGI is bypassed when people talk commercial Unix? This still mystifies me.
Worse, the evaluation placed high emphasis on admin tools. Around here, with Solaris, Irix, HP-UX, Digital Unix and Linux in one shop, noone is ever going to use SMIT or SAM or whatever. Just too much pain to learn them all.
Unifying sysadmin tasks is best done with scripting, at which Unix excels (as opposed to NT, say).
If your shop has one flavor of Unix only and has few, if any experienced sysadmins, then admin tools are for you. However, I don't think that most shops in the gov domain are small and hence emphasis on administration ease seem misguided.
What equally mystifies me is that the Compaq unit beats the competition 2:1 on CPU benchmarks, but only gets 3 stars on performance. Yes, it was weak in the graphics dept., but how can you loose so badly that you give up a 2:1 advantage? In my workstation I'd rather tolerate weakish graphics than a slow CPU. Of course, we are never told actual performance numbers and how they were weighted. Too bad.
The article quotes Alan Baratz of SUN, not someone inside AOL/Netscape. AFAIK, SUN does not have its hands significantly in the browser pie, and Baratz is not at the helm of the combined SUN/AOL effort either. So, that piece of journalism is just FUD. Besides, the way I read the netscape/mozilla license, the code can't be closed anymore.
Its funny how SUN people push their misguided Community license stuff. It just won't fly. It's just a cheap attempt at grabbing attention and get a few gullible souls to work for them for free.
We have had an opportunity to benchmark HP's new N4000 server with 8 440 MHz PA-8500 CPUs in them. The results in a nutshell:
The PA-8500 inside that server lives up to what its SPECfp promises. It is 1.5-2x faster than SGI Origin2000 with 250 MHz MIPS R10000. That's with computational chemistry / molecular modelling codes crunching real world data sets up to 4 way parallel.
So, in contrast to what the above poster guessed, SPECfp does not seem to be skewed when compared to real world performance. I'd like to see how the machine scales up to 8way. Soon we'll know that too.
This is nothing new really. It is exactly what our immune system is doing already, and very well at that. The immune system even continues to function when being hammered with myriads of problems all at the same time or when receiving a strong do not act signal. And the immune system is just a part of the continuous growth and turnover in our bodies which is usually referred to as life.
Without the continous healing, rebuilding and growth occurring in our bodies, you and I would cease to exist within days, if not quicker.
Our bodies are not static. They are the product of continuous buildup and removal of molecules, precisely orchestrated in space and time and self-organizing. As such, we resemble much more the seemingly static patterns seen in flowing water than a machine that is first built and then used. See nonlinear thermodynamics and chaos theory.
I just wonder if purposely built 'machines/computers' to 'enhance' these functions will really do so, or rather get in the way.
Keep in mind that the Ultra5 is the frontend of the E10000, e.g. the console from which to manage the E10000. It has not much work to do, just needs to be here and monitor some activity. It is NOT used as a server.
Being a scientist myself, I feel that E-biomed is a great idea. Finally someone of the establishment challenges age-old ideas and tries to explore new territory. E-publishing (not the kind of 'yeah, we also offer paid access to online versions' of the journal publishers) will curb the rising costs libraries and individual subscribers face, it will do away with the signing away of copyright to publishers AND it will significantly challenge the peer review system that we are all so fond of (or are we?). Let me tackle each one:
Cost: Libraries have complained for many years about the rising costs of print subscriptions. Many have been forced to reduce the number of journals they subscribe to, all the while the number of journals being published has actually risen. Reasons for rising costs are many, but one chief contributor is the pressure on scientists to publish more - publish or perish is the key word. As a consequence, research results are being broken down into 'least publishable units' and pushed out into papers.
E-publishing in the E-biomed sense, while not necessarily curbing the proliferation of published papers, will at least significantly cut the cost of publishing and accessing papers.
There is also a growing amount of data that cannot be easily published in paper form, nor does it fit established scientific databases such as Genbank. E-publishing lets us put this stuff out there and make accessible to anybody who wants it.
Copyyright: It has long irked me and many scientists I know that many publishers require me to sign a copyright transfer agreement which robs me of most rights to my own output. If I decide to reuse a figure in an overview article, I actually have to go and ask permission to do so. Granted, I ususally do get that permission, but it bugs me to no end that I have to jump through such hoops just so that I can get into a first tier journal. It was my creativity, planning, execution and insight that enabled the publication in the first place. The publisher's contribution, while not negligible, is relativeley minor. E-publishing, if done right, will let me retain copyright of all my works.
Peer review: As many other contributors to this discussion have pointed out, peer review guarantees quality of content. However, science being an endeavor of humans, peer review is fraught with perils. It is mostly an old-boy network and hence new people, new ideas, unorthodox approaches and new interpretations of old data have a very hard time getting past the reviewers. By its very nature peer review resists change. Furthermore, because 'peers' are also mostly competitors, and because the review process is confidential, the system invites fraud on the part of the reviewers. Be that stealing ideas or simply slamming a paper in order to slow down the authors, it is not in the spirit of furthering science, but definitely occurs quite often.
Now contrast peer review with the experience we have been able to gather with the pysics eprint archive. Eprints are not reviewed formally, but despite this the eprint archive is highly regarded in the field and is probably the most widely used dissemination vehicle. Why? The wide readership and instant feedback that is enabled by the system essentially guarantees that authors are very careful to not make fools of themselves. This is not an oligarchy of the establishment as in peer review, but rather democracy in action. I'll take that anyday over peer review.
Christoph Weber
Now what's really missing is free access to data collected by the state about us. Europeans have that, and it is scary how much BS needs to be corrected sometimes when an individual takes the time and effort to check. ;-)
Then again, I prefer for them to sell BS to companies rather than the truth about me
Not necessarily a bad thing. After all, they must have my written consent.
Also think about how many times you have handed over income info voluntarily to commercial entities already.
Having said all that, I still have nagging doubts that there will be slip-ups and just plain against-the-rules actions on the part of the responsible people.
Journalling filesystems slower than non-journalling ones? Not XFS!
XFS is actually faster than its non-journalling predecessor, EFS. I had both on the same hardware (low-end Indy). EFS itself wasn't bad. I thought is was faster than SUN's ufs on otherwise comparable hardware with the apps I used back then.
I'd never go back to an old-style FS if I had a choice. Glad to see a key piece of technology open sourced.
Correct, you have to use MPI or something similar to run a single app across a cluster. BUT: There are tons of applications in science where one essentially runs 100s or 1000s of smallish calculations while varying some input parameters or input data. For these, a Linux cluster is the perfect match.
We are currently planning a 1000 CPU cluster for such an application where not even a high-speed network is needed, just lots of cycles to get the job done before we grow old.
right on!
Linux needs a fair bit of work until it can run on 64 CPUs or more, handle very big files, directories and databases and such
interestingly, SGI has stated that they want to migrate such features from Unicos and Irix into Linux and put the result back into the open
it'll be very interesting to watch events over the next two years or so...