If you are really interested in getting the four nines you cite (or the 5 nines, aka sigma-6), the important thing is not that any one machine is running the whole time (that is, aiming for five nines on any one node is pointless.)
The trick is to try to have your system up and running optimally 99.999% of the time. This means recognizing that nodes will fail, building in redundancies, having a plan in place for maintenance, and having a plan to make sure the maintenance plan is followed.
You are correct in saying that this is not something you can buy just by picking the right brand. That said, I don't think increasing quantity is always the best path to quality.
This post is not in reference to a press release. Its a full length article in the Journal of Neurophysiology.
The editors would not have published it if it wasn't leading edge, accurate, and reproducable.
As for who is doing the hyping, the story's submitter had an MIT address. At least on the surface, I don't see any grounds for calling this 'academic astroturf.'
Thats an excellent justification for the use of the technology in question. I understand that the parent has laid out the groundwork, not judged it favorably or infavorably. It does explain why law enforcement is so keen to get their hands on it.
Here's one simple, easy to digest scenario in which it harmful to society. Understand that I'm not taking a position on it, merely posing a hypothetical situation.
The terrorist-graph program is successful. Law enforcement decides to tweak the algorithms and use it on organized crime. It meets success and becomes part of the standard set of tools. The individuals in charge of this program are dedicated and enlightened and have no desire to abuse the system.
Fast forward 10 years. Most of the people in charge, and all of the political leaders have changed. They've inherited these law enforcement programs. They use them to dig up political dirt or other mischievous, but relatively light weight abuses. It is deeply entrenched in the bureaucracy.
Another 10 years. Corruption is heavy in high level politics. All likely challengers are identified ahead of time and neutralized, either with planted evidence or coercion.
I'm not attempting to make a slippery slope arguement here. Two assertions I'm making are:
Policy is relevant in a context that goes beyond the current administration. The current people in charge might or might not have noble intentions, but the next ones can always be worse.
Some people will abuse whatever resources are at their disposal.
Yeah, but when was the last time the Executive and Legislative branches (including both houses of Congress) were controlled by the same party for 5 consecutive years?
Jack Welch (former Chairman and CEO of General Electric) came up with the idea of "One Two or Out", which was if you can't be number one or number two in a market, get out.
It seems to work for GE, and Apple. And a lot of others.
Blind contributions. That is, I can make any donation I like to a political candidate, under a certain and arbitrary amount. The catch is, the politician has no way of confirming who made the contribution.
I can endorse my candidate and support them. They know they are getting support, but don't necessarilly know where it will come from.
Of the people who are supporting this so we can bury them in correspondence?
Is it really that useful to bury them in correspondence?
For instance, Mike Doyle is my representative. I'm going to call his office and ask wtf? in an intelligent, polite manner. I'm going to express my distate for his politics, express that I won't vote for him again (and omit that I didn't vote for him the first time). I'm going to pass this on to the local linux user's group (people who are aware of the broadcast flag and likely to be vocal.)
I don't want to sound like a yahoo. I don't want to have my voice drowned out in a sea of mass emails from people outside of the district. I want to make it clear to him that his position is costing him real votes. I don't want to be lumped in with the lunatic fringe.
The government doesn't, and shouldn't, have an obligation to protect the well-being, let alone property, of any individual. If you think otherwise, you really need to do a little research.
A society in which there is no protection of property or well being of individuals is in anarchy, whether there is a body that calls themselves a government or not.
This wasn't a press conference, and the questions weren't from journalists.
The questions were from Microsoft's partners (ie, VAR's, stockholders, people who have large investments in MS products)
They aren't paid to ask questions, they're paid to implement solutions. They want to make sure that their chosen partner is doing their fair share.
Its a two way street -- the MS execs get real feedback from real customers who aren't afraid to call the shots the way they see them. The partners get some added measure of a pulse from the management of their very important business partner.
I've got my list of red flag words in place, but I've been having some pretty serious problems writing an effective filter... Has it been solved already? Googling for a comprehensive profanity dictionary at work is... tricky.
Yes, the solution is:
Do peer reviews.
There are other benefits, which you can read about in books by the likes of Kent Beck and Martin Fowler. But I'll bet that if you made it a policy to do reviews, and a policy that profanity would not be tolerated, it would clear up sooner than later.
Could it really be that much different than working for a large, faceless corporation? The same problem seems to crop up wherever one works--dealing with other people.
It can be different than working in the faceless corporation. You just have to be in the right group. I'm staff at a large urban university, and it is great.
First off, I get to use the athletic facilities. That means I can swim 2000 meters at lunch every day.
Two, its very laid back. I work with a lot of foreigners, and we have a joke about the lab being on "French Standard Time".
Three, the researchers are no-nonsense and committed. They're used to doing things for themselves, but very grateful if you can help them out.
Four, lots of comaraderie. They pull all nighters, I don't but I don't punch out early, either.
Five, great benefits. Tuition reimbursment. Free public transportation.
Six, good environment. I work in a secured area, but there are always interesting surgeons and brain-scientists around. Fun people.
Seven, you are expected to challenge yourself. Always good, and I work with some highly motivated, world class scientists.
Eight, you can do something you believe in, not just something that fills some economic niche.
Nine, job security. Nobody gets fired (though we all wish sometimes that people were).
Ten, you get to work with some cool toys. I won't go into specifics, but my lab (about 20 people) spends over $200,000 on technology to work with each year.
The bad side -- its a beauracracy like any large organization. BUT the resources are there if you have the patience to figure out which strings to pull.
The bottom line is a university job is like any other. Make sure you have a good boss, and that even if you don't align your goals with hers, you can sleep at night knowing your working toward that end.
Apple's main beef with IBM seems to be that they are not developing PPC's with enough quantity and innovation.
Why would Intel put more into PPC than IBM? If Apple is going to switch to Intel chips, they aren't doing it because they think Intel will be a great champion of PowerPC.
First of all, Intel is not going to do anything that could jeopardize the x86 line. Dilluting the market (by increasing the number of PowerPC's in production) will not help Intel, even if they're selling them.
Second, why would Intel be able to improve the PPC more quickly than IBM. Switching horses in the middle of the stream is not the solution. Intel would need to hire engineers to develop PPC. They would need to get managers and execs to take ownership. They would need to divert resources away from x86. Not going to happen.
The world security situation isn't really any different than it has been for the past thirty years, but now we finally have the political will to build systems to give us more control over people's lives. People are now willing to give up more of their freedom for perceived security, and we're going to take the opportunity to do it.
I think this hit the nail on the head.
The rest of it makes it seem like there is a definite malice behind the initiative. I do not see it that way. This is a multi-billion dollar deal. There are people who stand to make a lot of money if they get this contract. Potentially hundreds of millions of dollars. They're going to do everything they can to push it through.
And damn the costs, the big picture, or the correctness of such a project. This is someone's shot at the big money. On the government side, its just beauracrats who think their job is to spend money and listen to presentations from corporate big-shots.
Then there's the electorate. Unfortunately, half of the population has below average intelligence.
Overly rigid structure and formatting are not helpful. Often times I think programmers are so used to writing for compilers that they forget that humans are much better parsers.
Design documents should be easilly written and easilly updated. I prefer a text editor to something where I can have 'really nice' formatting. Its just easier and quicker, and that leads to the documents being more likely to be kept up to date.
If the docs aren't up to date, no matter how well written or well designed they are, they are misleading and unhelpful.
Well, yes, you are not expected to remain sitting in one place for more than 20 minutes. Get up, walk around. Live your life.
I have heard the same thing from my eye doctor.
These are guidelines. If you want to be healthy, do this...
Its like "Drink 2 liters of water every day." I'd probably be in better shape if I did, but I still don't.
And by the way, if your room is 19x19 and you're too fat and lazy to get up, roll your chair to the corner and look to the opposite corner. You'll be looking at something 26.8 feet away.
Not interested in Exchange or Sharepoint, already use SQL Server.
I've used 2000 Server, but not Small Business Server.
My experience with creating MSI's has not been good. In short, I haven't figured out how to do anything useful with the domain policy, other than automatically install and setup Office (neat trick, but I want more), not allow users to turn off their computers, require them to change their passwords, and set IE's proxy to 0.0.0.0.
I had ideas about creating Windows installer packages for everything we use and controlling the distribution via policy, but the flexibility was not there. I'm not opposed to hiring an admin, but I'm not convinced that the payoff is really there.
That may work well for 5000 machines, but lets say I only have a couple. It scales up pretty well, but it doesn't scale down.
For 10 machines that turns into a lot of running around running Windows Update or the purchasing of an expensive server license.
For 50 machines, one person is going to spend a lot of time keeping the server oiled and making sure all the GPO's and software policies are up to date. It may not be any harder to administer 5000 machines than 50, but its an unreasonably high cost-per-cpu for the administration of the 50.
A good solution is this: Indoctrinate your users not to use Internet Explorer or Outlook. You'd be surprised how many spyware problems disappear.
Yes, you can GPO Internet Explorer to use a 0.0.0.0 proxy, but then you break Windows Update.
What I'm saying is, for small networks, GPO's and AD's and server licenses are a big time sink.
Probably because the little guy is already locked in, Microsoft doesn't put much effort into making a better experience for them. They are really trying to get into the server room though, so their products are geared toward that.
"Would I build a supercomputer out of them [opterons]? No... If you are building new systems odds are you are going to go PPC."
http://www.cray.com/products/xt3/index.html
The link above goes to pretty much the top of the line Cray. It is built out of Opterons and PPC. The PPC is the DMA controller, the Opteron actually executes your code.
So maybe you wouldn't choose Opterons, but Cray would.
Slashdot Mirror
If you are really interested in getting the four nines you cite (or the 5 nines, aka sigma-6), the important thing is not that any one machine is running the whole time (that is, aiming for five nines on any one node is pointless.)
The trick is to try to have your system up and running optimally 99.999% of the time. This means recognizing that nodes will fail, building in redundancies, having a plan in place for maintenance, and having a plan to make sure the maintenance plan is followed.
You are correct in saying that this is not something you can buy just by picking the right brand. That said, I don't think increasing quantity is always the best path to quality.
This post is not in reference to a press release. Its a full length article in the Journal of Neurophysiology.
The editors would not have published it if it wasn't leading edge, accurate, and reproducable.
As for who is doing the hyping, the story's submitter had an MIT address. At least on the surface, I don't see any grounds for calling this 'academic astroturf.'
Here's one simple, easy to digest scenario in which it harmful to society. Understand that I'm not taking a position on it, merely posing a hypothetical situation.
The terrorist-graph program is successful. Law enforcement decides to tweak the algorithms and use it on organized crime. It meets success and becomes part of the standard set of tools. The individuals in charge of this program are dedicated and enlightened and have no desire to abuse the system.
Fast forward 10 years. Most of the people in charge, and all of the political leaders have changed. They've inherited these law enforcement programs. They use them to dig up political dirt or other mischievous, but relatively light weight abuses. It is deeply entrenched in the bureaucracy.
Another 10 years. Corruption is heavy in high level politics. All likely challengers are identified ahead of time and neutralized, either with planted evidence or coercion.
I'm not attempting to make a slippery slope arguement here. Two assertions I'm making are:
Yeah, but when was the last time the Executive and Legislative branches (including both houses of Congress) were controlled by the same party for 5 consecutive years?
Hey, this is Slashdot. Banning Windows is always a popular idea!
(How MS gets dragged into an environmental debate, well...)
Jack Welch (former Chairman and CEO of General Electric) came up with the idea of "One Two or Out", which was if you can't be number one or number two in a market, get out.
It seems to work for GE, and Apple. And a lot of others.
The stars say that this project is going to come in overbudget!
How about a third option:
Blind contributions. That is, I can make any donation I like to a political candidate, under a certain and arbitrary amount. The catch is, the politician has no way of confirming who made the contribution.
I can endorse my candidate and support them. They know they are getting support, but don't necessarilly know where it will come from.
Food for thought.
Is it really that useful to bury them in correspondence?
For instance, Mike Doyle is my representative. I'm going to call his office and ask wtf? in an intelligent, polite manner. I'm going to express my distate for his politics, express that I won't vote for him again (and omit that I didn't vote for him the first time). I'm going to pass this on to the local linux user's group (people who are aware of the broadcast flag and likely to be vocal.)
I don't want to sound like a yahoo. I don't want to have my voice drowned out in a sea of mass emails from people outside of the district. I want to make it clear to him that his position is costing him real votes. I don't want to be lumped in with the lunatic fringe.
A society in which there is no protection of property or well being of individuals is in anarchy, whether there is a body that calls themselves a government or not.
Lets be clear about this: anarchy is bad.
Its a bit like nuclear weapons -- You do not have to use them to serve a purpose. The threat of eradicating your enemies is quite powerful.
Read the article.
This wasn't a press conference, and the questions weren't from journalists.
The questions were from Microsoft's partners (ie, VAR's, stockholders, people who have large investments in MS products)
They aren't paid to ask questions, they're paid to implement solutions. They want to make sure that their chosen partner is doing their fair share.
Its a two way street -- the MS execs get real feedback from real customers who aren't afraid to call the shots the way they see them. The partners get some added measure of a pulse from the management of their very important business partner.
So what is it about this situation that you can't do code reviews? I think its absolutely one of the best ways to crank up the quality of your output.
After a surprisingly short time, people just start writing the kind of code that the team agrees is the right kind.
Don't knock it till you try it.
Yes, the solution is:
Do peer reviews.
There are other benefits, which you can read about in books by the likes of Kent Beck and Martin Fowler. But I'll bet that if you made it a policy to do reviews, and a policy that profanity would not be tolerated, it would clear up sooner than later.
It can be different than working in the faceless corporation. You just have to be in the right group. I'm staff at a large urban university, and it is great.
First off, I get to use the athletic facilities. That means I can swim 2000 meters at lunch every day.
Two, its very laid back. I work with a lot of foreigners, and we have a joke about the lab being on "French Standard Time".
Three, the researchers are no-nonsense and committed. They're used to doing things for themselves, but very grateful if you can help them out.
Four, lots of comaraderie. They pull all nighters, I don't but I don't punch out early, either.
Five, great benefits. Tuition reimbursment. Free public transportation.
Six, good environment. I work in a secured area, but there are always interesting surgeons and brain-scientists around. Fun people.
Seven, you are expected to challenge yourself. Always good, and I work with some highly motivated, world class scientists.
Eight, you can do something you believe in, not just something that fills some economic niche.
Nine, job security. Nobody gets fired (though we all wish sometimes that people were).
Ten, you get to work with some cool toys. I won't go into specifics, but my lab (about 20 people) spends over $200,000 on technology to work with each year.
The bad side -- its a beauracracy like any large organization. BUT the resources are there if you have the patience to figure out which strings to pull.
The bottom line is a university job is like any other. Make sure you have a good boss, and that even if you don't align your goals with hers, you can sleep at night knowing your working toward that end.
Apple's main beef with IBM seems to be that they are not developing PPC's with enough quantity and innovation.
Why would Intel put more into PPC than IBM? If Apple is going to switch to Intel chips, they aren't doing it because they think Intel will be a great champion of PowerPC.
First of all, Intel is not going to do anything that could jeopardize the x86 line. Dilluting the market (by increasing the number of PowerPC's in production) will not help Intel, even if they're selling them.
Second, why would Intel be able to improve the PPC more quickly than IBM. Switching horses in the middle of the stream is not the solution. Intel would need to hire engineers to develop PPC. They would need to get managers and execs to take ownership. They would need to divert resources away from x86. Not going to happen.
If Apple now owns a large amount of PPC IP, why would they switch to Intel?
I can't think of a worse way to recoup on that intellectual property then to abandon PPC.
I think this hit the nail on the head.
The rest of it makes it seem like there is a definite malice behind the initiative. I do not see it that way. This is a multi-billion dollar deal. There are people who stand to make a lot of money if they get this contract. Potentially hundreds of millions of dollars. They're going to do everything they can to push it through.
And damn the costs, the big picture, or the correctness of such a project. This is someone's shot at the big money. On the government side, its just beauracrats who think their job is to spend money and listen to presentations from corporate big-shots.
Then there's the electorate. Unfortunately, half of the population has below average intelligence.
Lots of non-former-Baby Bell ISP's will give you a DSL line without a telephone line.
Shop around.
Overly rigid structure and formatting are not helpful. Often times I think programmers are so used to writing for compilers that they forget that humans are much better parsers.
Design documents should be easilly written and easilly updated. I prefer a text editor to something where I can have 'really nice' formatting. Its just easier and quicker, and that leads to the documents being more likely to be kept up to date.
If the docs aren't up to date, no matter how well written or well designed they are, they are misleading and unhelpful.
Well, yes, you are not expected to remain sitting in one place for more than 20 minutes. Get up, walk around. Live your life.
I have heard the same thing from my eye doctor.
These are guidelines. If you want to be healthy, do this...
Its like "Drink 2 liters of water every day." I'd probably be in better shape if I did, but I still don't.
And by the way, if your room is 19x19 and you're too fat and lazy to get up, roll your chair to the corner and look to the opposite corner. You'll be looking at something 26.8 feet away.
Not interested in Exchange or Sharepoint, already use SQL Server.
I've used 2000 Server, but not Small Business Server.
My experience with creating MSI's has not been good. In short, I haven't figured out how to do anything useful with the domain policy, other than automatically install and setup Office (neat trick, but I want more), not allow users to turn off their computers, require them to change their passwords, and set IE's proxy to 0.0.0.0.
I had ideas about creating Windows installer packages for everything we use and controlling the distribution via policy, but the flexibility was not there. I'm not opposed to hiring an admin, but I'm not convinced that the payoff is really there.
That may work well for 5000 machines, but lets say I only have a couple. It scales up pretty well, but it doesn't scale down.
For 10 machines that turns into a lot of running around running Windows Update or the purchasing of an expensive server license.
For 50 machines, one person is going to spend a lot of time keeping the server oiled and making sure all the GPO's and software policies are up to date. It may not be any harder to administer 5000 machines than 50, but its an unreasonably high cost-per-cpu for the administration of the 50.
A good solution is this: Indoctrinate your users not to use Internet Explorer or Outlook. You'd be surprised how many spyware problems disappear.
Yes, you can GPO Internet Explorer to use a 0.0.0.0 proxy, but then you break Windows Update.
What I'm saying is, for small networks, GPO's and AD's and server licenses are a big time sink.
Probably because the little guy is already locked in, Microsoft doesn't put much effort into making a better experience for them. They are really trying to get into the server room though, so their products are geared toward that.
"Would I build a supercomputer out of them [opterons]? No... If you are building new systems odds are you are going to go PPC."
http://www.cray.com/products/xt3/index.html
The link above goes to pretty much the top of the line Cray. It is built out of Opterons and PPC. The PPC is the DMA controller, the Opteron actually executes your code.
So maybe you wouldn't choose Opterons, but Cray would.
If I believed in what they are doing over there, I'd be over there, why aren't you?