you have far more problems than not understanding single vs. dual cpu systems when it comes to a server environment
The cheapest and easiest way to scale a dynamic web site up to handling a shit-ton of load is to load balance across multiple web servers. It's not to add more CPUs to the current web server. Ditto with database servers, although most DB vendors make that kind of clustering expensive enough that customers find it easier to upgrade their existing server.
That's not true for every kind of server out there, but it's true for web servers.
They are contending for PCI bus bandwidth, disk controller bandwidth, and (like I said before) memory bandwidth. Either your needs are lightweight enough that storing your database and your web pages on the same disk are basically fine, or your needs are heavyweight enough that you'll get better performance for less by separating out the systems further.
(and by "everything else", I mean the integration that a distribution brings -- how well are menus configured in your chosen desktop environment, does it have a good package installation story that keeps those menus up to date, does it provide you with recent and stable versions of popular software, etc)
Actually, I'd be more interested in other things- how well the selection of packages and the versions thereof meshes with stated goals, the impact of distro patches to the software, that kind of thing. For Slackware, which tries to be familiar to old UNIX hands, it seems reasonable to evaluate how well it mirrors the experience of running e.g. Solaris.
A pox upon the developers who go after that.3% performance increase before they've improved their algorithmic efficiency, before they've fixed all memory leaks, before they've figured out how to handle a responsive UI without relying on that mythical and imperceptible performance boost, and before they've figured out how to change the code that implements that performance boost without introducing more bugs.
Unless you've worked on a system that requires it, shut the hell up about what you 'should' do in an embedded environment. The Linux kernel runs in many embedded spaces these days, and lightweight it is not. If you look at real-time Linux, for instance, they didn't make it 'real time' by tuning the crap out of array-vs-pointer management. No, they fixed the design so that it didn't matter.
Indeed, that sort of functionality is why I use graphical clients at all. If I didn't mind seeing only one message or one folder at a time, I'd still be using pine (and trn for news, for that matter).
However, you're mistaken if you think major clients can't handle that. You can disable the message pane in OS X Mail, Outlook, or Thunderbird so that all messages open in new windows (even without doing so, you can double click on any message to open it in a new window); in at least Thunderbird, you can double click on a folder to open it as a new window as well (I haven't tried it in Outlook, and I don't think it's possible in OS X Mail).
All three of those clients also support offline mode, and do it quite well I might add.
"They never would have been announced during 2004 had Microsoft not first revealed that it was making the feature a standard feature of the next Windows."
This is patently false; Apple hired Dominic Giampaolo, developer of BeFS (which was specifically developed to have the sort of 'fast search' that is finally showing up in mainstream operating systems), in February of 2002. The intent was clear, back in 2002, that it was Apple's intent to bring the innovations of BeFS to OS X, a year before Microsoft announced the feature.
Phrasing the chain of events as "When Microsoft announced [it] in October 2003, the race began." is ridiculous. Apple effectively announced the plan 18 months prior, and even then it was clear that it was too late to make it into 10.2, the 10.3 release was unlikely, and that therefore... it would show up in 10.4. Just like it did.
More damning, though, is that Microsoft has announced this feature a number of times, every time they've announced that a future OS (starting with NT 5, IIRC) would feature a database-driven filesystem. Why didn't anyone else jump on getting the feature first then, rather than this time? I'll tell you why: it's a hard feature that took a lot of time to work on, and every one had been working on it the whole time.
The real problem here, though, is that I bet Paul Thurrott doesn't know any of this. All he knows is, Spotlight Search was announced when 10.4 was announced, which was after Microsoft announced it. And without looking at it any closer, he decided he knew the whole story and that he could speak authoritatively on the subject. I can't be bothered to read the rest of the article if it has the same empty authoritative voice.
This is the site that last week, had an op-ed up arguing that "loving" Microsoft is OK, and Linux is just the product of some nefarious cabal of hypesters and PR men. Yeah, uh, I don't see me caring about this review of inkjet printers either. One of the things that matters to me is whether I can print to it in Linux, which I kind of doubt they'll be able to handle.
Interestingly linked questions. My experience has been that good programmers have a mix of intelligence, theoretical knowledge, and practical experience; a good degree program can provide those. A lot of other things, like reading and programming, can provide them too. Further, a programmer can learn more and get better, and gain more experience and get better; it's hard to articulate whether a programmer can get 'smarter' or not.
My own experience is that the best kind of experience involves dealing with the conflict between time, budget, and design; choosing to ignore any of them (or being able to) makes the experience less valuable in terms of how much you learn from it.
Ummm... what do you know about the reasons behind Apple's strategy, that you can precisely enumerate the reasons?
What makes you think Intel's DRM, and not AMD's supply problems, is the primary reason Apple is skipping AMD? Do you really think that if AMD had DRM capability in its chips, and Intel did not, that Apple would pick AMD?
It would be fairly trivial to approximate something more complex like "second saturday" by calling a script once a Saturday, and let it apply arbitrary checks to the date before sending the reminder. A cron job is cheap, the outside-cron check is cheap.
Perhaps you are comparing high-end CRTs to the LCDs that could be bought a few years ago? If you can get "high resolution" CRTs without eyestrain... I must assume you're talking about at least 1600x1200 at 85Hz or more. That's a damned expensive CRT, and until recently you had to pay about that much (if not more) to get an entry-level LCD doing 1024x768.
Modern LCDs have improved significantly... The worst LCDs of today outstrip most LCDs of a couple years ago. A good LCD that does 1600x1200 and doesn't have ghosting problems will run under 600USD. That's a difference of about 150USD and between 30 and 50 pounds. Make that two, and you're looking at as much as 100 extra pounds sitting on your desk.
With all that said... this DoubleSight doesn't look like a very good deal. 19" LCDs run around 300USD, buying two of them leaves you with a budget of 300USD on your mounting system (more than enough these days) and you'll still save money.
Re:Throw all the PR people and lawyers in the ocea
on
Paul Graham on PR
·
· Score: 1
No, a big chunk of PR isn't even that - it's just encouraging a particular (true) story to get out. RTFA.
Perhaps I'm misunderstanding you, but yes, I believe Irix supports this for its ccNUMA machines, where the 'distance' between CPUs (and associated memory) can vary quite a bit. If you've got a single system image running on 10 machines with 2 CPUs apiece, you really don't want it to treat every CPU as adjacent to every memory area.
I should probably look at all-in-ones again soon, but my experience last time around was that I can't stand them.
I want to only have to carry around one device... but when I'm wanting to listen to music, I don't want to deal with a stylus or phone buttons... I want the iPod's wheel (and storage). When I'm entering notes, I don't want the iPod's wheel or a thumb-sized keyboard, I want a pen-like device (unless it's reasonably close to full-sized, I'll write it out faster than I can type it out). When I want to talk on the phone, I don't want a big screen pressed up against my face... and I don't want the bulk of a PDA in my pocket whenever I want to carry my phone.
The net result is that I have three devices, and depending on circumstances I might carry 1, 2, or 3 of these devices at a time. I have to get creative with pockets, wearing a jacket, and so on... but the pay off is that everything really is quite easy to use. Using iSync to keep my address book straight across the phone and PDA and computer, and Bluetooth when I want to dial a number that isn't synchronized (uncommonly used numbers), I get along pretty well. It's not ideal, but neither is trying to use a smart phone to hold 150 albums of music, or make a PDA be a decent phone.
In addition to all that, I want Bluetooth in my phone: OS X's AddressBook is pretty slick with Bluetooth, never mind wireless synchronization and headsets. OS X-supported, Bluetooth-capable smart phones are in pretty short supply right now:-)
Software isn't its own thing, the software is part of the cost and benefit - something 'you PC people' often forget. But then, you're the one who said "A mere PC with comparable software stats can be had..."
Honestly? A $500 Mac mini isn't worth it to install OpenBSD or Linux on... because you're spending a couple hundred dollars on software you won't use, and hardware features you can't. So what? That's a given with any Mac hardware - you're buying it because you want the benefits provided by Mac software.
Which open source movie NLE can handle HD? Kino? Cinelerra? Which open source package matches any of the ones I listed for usability, power, or features? If you wanna go pure open source for these things, you've got to adopt RMS' approach of freedom first, features second.
I think you'd be hard-pressed to spend an additional $500 for you $199 PC and get movie-editing software that can handle HDV, multi-track audio content creation with musical notation, decent multimedia content organization, decent PIM software, and Quicken.
Slashdot Mirror
you have far more problems than not understanding single vs. dual cpu systems when it comes to a server environment
The cheapest and easiest way to scale a dynamic web site up to handling a shit-ton of load is to load balance across multiple web servers. It's not to add more CPUs to the current web server. Ditto with database servers, although most DB vendors make that kind of clustering expensive enough that customers find it easier to upgrade their existing server.
That's not true for every kind of server out there, but it's true for web servers.
They aren't selling them to be combined web+database servers.
That doesn't matter.
They are contending for PCI bus bandwidth, disk controller bandwidth, and (like I said before) memory bandwidth. Either your needs are lightweight enough that storing your database and your web pages on the same disk are basically fine, or your needs are heavyweight enough that you'll get better performance for less by separating out the systems further.
What's wrong is that the database and web server are probably not contending for CPU time anyway. They are both contending for disk and memory access.
But damn it, it's "furor" not "ferver." Nor should "furor" be confused wyth "Führer."
Actually, I'd be more interested in other things- how well the selection of packages and the versions thereof meshes with stated goals, the impact of distro patches to the software, that kind of thing. For Slackware, which tries to be familiar to old UNIX hands, it seems reasonable to evaluate how well it mirrors the experience of running e.g. Solaris.
A pox upon the developers who go after that .3% performance increase before they've improved their algorithmic efficiency, before they've fixed all memory leaks, before they've figured out how to handle a responsive UI without relying on that mythical and imperceptible performance boost, and before they've figured out how to change the code that implements that performance boost without introducing more bugs.
Unless you've worked on a system that requires it, shut the hell up about what you 'should' do in an embedded environment. The Linux kernel runs in many embedded spaces these days, and lightweight it is not. If you look at real-time Linux, for instance, they didn't make it 'real time' by tuning the crap out of array-vs-pointer management. No, they fixed the design so that it didn't matter.
Indeed, that sort of functionality is why I use graphical clients at all. If I didn't mind seeing only one message or one folder at a time, I'd still be using pine (and trn for news, for that matter).
However, you're mistaken if you think major clients can't handle that. You can disable the message pane in OS X Mail, Outlook, or Thunderbird so that all messages open in new windows (even without doing so, you can double click on any message to open it in a new window); in at least Thunderbird, you can double click on a folder to open it as a new window as well (I haven't tried it in Outlook, and I don't think it's possible in OS X Mail).
All three of those clients also support offline mode, and do it quite well I might add.
"Post-gres"
"They never would have been announced during 2004 had Microsoft not first revealed that it was making the feature a standard feature of the next Windows."
This is patently false; Apple hired Dominic Giampaolo, developer of BeFS (which was specifically developed to have the sort of 'fast search' that is finally showing up in mainstream operating systems), in February of 2002. The intent was clear, back in 2002, that it was Apple's intent to bring the innovations of BeFS to OS X, a year before Microsoft announced the feature.
Phrasing the chain of events as "When Microsoft announced [it] in October 2003, the race began." is ridiculous. Apple effectively announced the plan 18 months prior, and even then it was clear that it was too late to make it into 10.2, the 10.3 release was unlikely, and that therefore... it would show up in 10.4. Just like it did.
More damning, though, is that Microsoft has announced this feature a number of times, every time they've announced that a future OS (starting with NT 5, IIRC) would feature a database-driven filesystem. Why didn't anyone else jump on getting the feature first then, rather than this time? I'll tell you why: it's a hard feature that took a lot of time to work on, and every one had been working on it the whole time.
The real problem here, though, is that I bet Paul Thurrott doesn't know any of this. All he knows is, Spotlight Search was announced when 10.4 was announced, which was after Microsoft announced it. And without looking at it any closer, he decided he knew the whole story and that he could speak authoritatively on the subject. I can't be bothered to read the rest of the article if it has the same empty authoritative voice.
This is the site that last week, had an op-ed up arguing that "loving" Microsoft is OK, and Linux is just the product of some nefarious cabal of hypesters and PR men. Yeah, uh, I don't see me caring about this review of inkjet printers either. One of the things that matters to me is whether I can print to it in Linux, which I kind of doubt they'll be able to handle.
Add 'CoolTechZone' to the list of sites that feed trolls. That article manages to make OS Views look well informed and insightful.
Interestingly linked questions. My experience has been that good programmers have a mix of intelligence, theoretical knowledge, and practical experience; a good degree program can provide those. A lot of other things, like reading and programming, can provide them too. Further, a programmer can learn more and get better, and gain more experience and get better; it's hard to articulate whether a programmer can get 'smarter' or not.
My own experience is that the best kind of experience involves dealing with the conflict between time, budget, and design; choosing to ignore any of them (or being able to) makes the experience less valuable in terms of how much you learn from it.
Ummm... what do you know about the reasons behind Apple's strategy, that you can precisely enumerate the reasons?
What makes you think Intel's DRM, and not AMD's supply problems, is the primary reason Apple is skipping AMD? Do you really think that if AMD had DRM capability in its chips, and Intel did not, that Apple would pick AMD?
"Insightful" my ass.
It would be fairly trivial to approximate something more complex like "second saturday" by calling a script once a Saturday, and let it apply arbitrary checks to the date before sending the reminder. A cron job is cheap, the outside-cron check is cheap.
Perhaps you are comparing high-end CRTs to the LCDs that could be bought a few years ago? If you can get "high resolution" CRTs without eyestrain... I must assume you're talking about at least 1600x1200 at 85Hz or more. That's a damned expensive CRT, and until recently you had to pay about that much (if not more) to get an entry-level LCD doing 1024x768.
Modern LCDs have improved significantly... The worst LCDs of today outstrip most LCDs of a couple years ago. A good LCD that does 1600x1200 and doesn't have ghosting problems will run under 600USD. That's a difference of about 150USD and between 30 and 50 pounds. Make that two, and you're looking at as much as 100 extra pounds sitting on your desk.
With all that said... this DoubleSight doesn't look like a very good deal. 19" LCDs run around 300USD, buying two of them leaves you with a budget of 300USD on your mounting system (more than enough these days) and you'll still save money.
No, a big chunk of PR isn't even that - it's just encouraging a particular (true) story to get out. RTFA.
Uh... the difference is obvious on my $20 Sony headphones. 128kbps just sucks.
"Many, many CDs"? Hahahahahahano.
You're just feeling doomed and gloomy, bud.
Buy the friggin' CD. Then, you own it. Problem solved; that hasn't ever not worked for me.
Perhaps I'm misunderstanding you, but yes, I believe Irix supports this for its ccNUMA machines, where the 'distance' between CPUs (and associated memory) can vary quite a bit. If you've got a single system image running on 10 machines with 2 CPUs apiece, you really don't want it to treat every CPU as adjacent to every memory area.
I should probably look at all-in-ones again soon, but my experience last time around was that I can't stand them.
I want to only have to carry around one device... but when I'm wanting to listen to music, I don't want to deal with a stylus or phone buttons... I want the iPod's wheel (and storage). When I'm entering notes, I don't want the iPod's wheel or a thumb-sized keyboard, I want a pen-like device (unless it's reasonably close to full-sized, I'll write it out faster than I can type it out). When I want to talk on the phone, I don't want a big screen pressed up against my face... and I don't want the bulk of a PDA in my pocket whenever I want to carry my phone.
The net result is that I have three devices, and depending on circumstances I might carry 1, 2, or 3 of these devices at a time. I have to get creative with pockets, wearing a jacket, and so on... but the pay off is that everything really is quite easy to use. Using iSync to keep my address book straight across the phone and PDA and computer, and Bluetooth when I want to dial a number that isn't synchronized (uncommonly used numbers), I get along pretty well. It's not ideal, but neither is trying to use a smart phone to hold 150 albums of music, or make a PDA be a decent phone.
In addition to all that, I want Bluetooth in my phone: OS X's AddressBook is pretty slick with Bluetooth, never mind wireless synchronization and headsets. OS X-supported, Bluetooth-capable smart phones are in pretty short supply right now :-)
Software isn't its own thing, the software is part of the cost and benefit - something 'you PC people' often forget. But then, you're the one who said "A mere PC with comparable software stats can be had..."
Honestly? A $500 Mac mini isn't worth it to install OpenBSD or Linux on... because you're spending a couple hundred dollars on software you won't use, and hardware features you can't. So what? That's a given with any Mac hardware - you're buying it because you want the benefits provided by Mac software.
Hahahahahahahaha
Which open source movie NLE can handle HD? Kino? Cinelerra? Which open source package matches any of the ones I listed for usability, power, or features? If you wanna go pure open source for these things, you've got to adopt RMS' approach of freedom first, features second.
Which isn't bad, but be honest about it, OK?
Comparable software stats?
You mean, something that includes something comparable to iMovie HD, iDVD, iPhoto, and Garage Band, OS X Mail, iChat, Address Book, iCal, iSync, and Quicken 2005?
I think you'd be hard-pressed to spend an additional $500 for you $199 PC and get movie-editing software that can handle HDV, multi-track audio content creation with musical notation, decent multimedia content organization, decent PIM software, and Quicken.