When someone suggests Dia as an alternative to Visio, I'll laugh. Ahah:) Even Smartdraw, a somewhat easier-to-use-and-a-lot-cheaper alternative to Visio, pales in comparison. It's a bit like suggesting ms paint for cad work.
1) color management and color management support by application software
2) screen real estate - in windows, I get much more workable screen area at the same dpi than any Xorg configuration. I've installed recently windows 8 preview on a XM, booted at 800x600 and no widget, button or whatever was out-of-screen
3) software UI consistency - every windows program looks the same, and the options are at a predictable place; the ribbon UI breaks this, but only until every other program has a ribbon (I hate it)
4) decent clipboard - if I copy something from an application and paste it on another application, I really don't care about what widget toolkit app A is using and if B can decode the format. Vector, image and formatted text copy/paste in xorg are a nightmare
5) stability - windows crashes more than my bsd servers, but every x/xorg I've tried is much more unstable than any windows I've used since NT3.5. On a *nix desktop, having X crashing is almost the same as rebooting
6) decent font support. Windows comes with some nice system fonts, and very good TTF support. Font rendering in windows has some nifty algorithms thay maybe many don't like it, but I do - I can distinguish between semi-bold and bold text
7) explorer performance - try to open a folder with 500 hi-res pictures and wait for the previews in gnome. Now imagine it if it was 2 thousand pictures
8) consistency - in windows, you can navigate trough menus and usually find the options you expect on the place you are looking
9) software variety - you have virtually everything for windows
10) ease of administration - the granularity of the administration knobs (local policy or domain policy) is something almost unheard on *nix destop distros
11) consistent upgrade path - there's a video somewhere of somewone who upgraded windows from windows 1.0 to vista and documented the process. I usually don't upgrade windows installations - I prefer to rotate the disk as a backup and do a clean install - but it is nice to have options.
12) print spooler and print support - windows has ONE print spooler and very good usermode driver print support. That translate to not having to use print processors and interpreters, LPD and CUPS, and subpar generic drivers. There are exceptions - and I even somewhat like CUPS, even if now is an Apple product - but the ppd drivers are usually not that impressive, even when supplied from the manufacturer
13) RDP(v5) - there is no competing technology at the same price. Yes I've tried FreeNX. And VNC. And X tunnedled trough ssh. FreeNX (or whatever was called previously) is quite good, but still no match for RDP
14) multimedia - I don't need Jack funcionality, I need to play multiple sound streams on the same computer without chmod voodoo or to know about audio toolkits. Just need to quickly play a video with my favourite player without needing to stop my mp3 player
15) multiple monitor support. It just works. and you can rotate the screen, or send it to a projector via wireless.
Many/all of these points are addressed by other operating systems, such as MacOS X, BeOS/Haiku (go Haiku!), etc. I've been using windows on desktop since 3.0, so my list is probably biased, because I "hate" x/xorg with passion. Btw, I'm also an unix professional, and use BSD (FreeBSD/OpenBSD) as a part of my daily job, and I've been trying to convince myself to use x/xorg without success since 1995.
Apparently the numbers were only high enough to cause an uproar after the start of persecution of jews in Germany, some year before WWII.
It is easy to have a prosperous nation when you have wealth (from all those fleeding immigrants) and access to unlimited credit. In many regions of the world minorities are slaughtered before they can afford to flee the country they live in, so yes, I'm talking about jewish wealth. If I was one of the minorities being persecuted in Germany and had the choice to leave the country, I'd do it in a hearbeat. But while the jews are no different from other minorities, such as tutsi in Rwanda, the victims of serbian ethnic cleansing, the curds in Iraq (and during Allied supervision of the post gulf war), etc.
Israel allies don't really care for justice or amend of past mistakes - they only care about money. They are the same that overthrown gadaffi -a merciless dictactor, but with a stable country - but in Syria the bodies keep piling and they don't care. If Syria was a major oil and gas producer, they'd sing a different tune. Or maybe if they were jews, they could aspire to have first-class treatment.
Thats what happens when outsiders decide to create a country where some other outsiders live, without regard for the people that actually live there, and their neighbours. If the WWII never happened (and the holocaust), do you believe Israel would exist?
Except that they also use the "self-defense" excuse to attack without military provocation. Can your colleagues punch you in the face now, because someday you may be an asshole to them? Tought so.
Given the military capability of Israel, usually their self-defense is like shooting the kids who threw rocks at your window with a laser-guided missile.
So, because they have been kicked around and harassed, they have the right to do as they please? And how convenient of you to omit the influx of jews to what now is Israel after they assisted Britain in taking control of the territory in 1917. But when the arabic population revolted against what could have been seen as an "invasion", and the brits imposed restrictions on jewish immigration, they turned on them. During the post-WWII era, the palestinian territories were flooded by holocaust survivors.
Now think about this - pick a nationality you don't like very much and/or has different values than yours, imagine that suddenly your whole neighbourhood is populated by those guys, and when you complain they tell you "god gave me this land". You call the authorities and the court decides "ok now your whole neighbourhood is a different nation where their law and values are followed - not yours". Would you accept it?
It is hard to be sympathetic to the Israeli position- they are a nuclear nation, they have as allies some of the more powerful countries in the world, and israeli jewish interests have a lot of weight in international politics - and not necessarily those of the muslim population who also are israeli citizens. Also, the hipocrisy of some Israeli allies is staggering - I'm shure the american right wing would be delighted to have a new nation forced by the UN on their territory, called for example "new new mexico", or "chinazaquistan", as a way to solve their immigration problems.
I'm not defending the extremism against Israel taken by their neighbours (and religiously I really don't care about muslim or jewish deities, or any kind of deities), but one must be blind to not see the manipulation and hipocrisy of their government, and their seemingly overwelming political and economical influence.
I think you are talking about java and eclipse development tools that reportedly run faster using Linux ABI in FreeBSD than in native linux. If I recall correctly, at the time, the mainstream linux kernel was the 2.4 line, that had several design issues that impacted performance on many applications. 2.6 is a whole different beast. The Linux ABI is a nice FreeBSD feature (and currently is 2.6 compatible), but there is no support for 64 bit binaries. Also, some syscalls are not implemented, as well as some exotic functions such as LDT manipulation and virtual 86 mode. The biggest gripe I get with the Linux ABI are the userland environments available on the ports tree - some are old, some are deprecated and/or forbidden. It would be awesome to fire up a jail running a linux userland to use linux-only software, even if limited to 32 bit - Zend Encoder comes to mind.
He can build the package on a separate machine, with the fixed dependencies, and then install it on the VMs via puppet. Package building on FreeBSD isn't as simple as it should be (I prefer the simplicity of the OpenBSD ports in this regard, but I don't recall any broken dependency incident when building PHP. It is not a task I do everyday - once a month or so, but never had any issues as described.
I am a FreebSD user, and most of my work envolves BSD setups.
For most of what I've benchmarked, CentOS 5.4 is faster out-of-the-box than a tuned FreeBSD install. Upon resurce starvation, I find FreeBSD much more stable and easier to recover to a running state. Also, tools like zfs and jails allow me to be able to do more with the machine, and even improve the maintainability of the system. The ports infrastructure is not perfect, but does a good job at it, and I often get less garbage when upgrading packages in freebsd than on linux.
The stability argument is moot nowadays, as I find a CentOS or Debian install as stable as FreeBSD (and forgeting the crappy unstable mess that were the 5.x releases. Most of the arguments against linux or freebsd used in the past arent valid today -as an example, linux has a ton of crappy drivers, freebsd still has a ton of unsupported hardware. Regarding 3D acceleration linux is somewhat far ahead, but behind other operating systems.
I don't play around much with DragonFly, but I think the direction they're heading is grid computing, and not necessarily SMP. The project implements some interesting concepts, and are one of the few operating systems actually trying something "new", and for that they deserve all the merit they can get.
AFAIK most of the top 500 supercomputers run a heavily modified and customized version of linux, not your run-to-the-mill distro. Given that most BSDs were late to the "SMP race" and that Linux already supports a set of architectures at least as vast as netbsd, it is the ideal candidate to be tewaked and installed on those supercomputers. If by supercomputers you ment clusters, yes, there is a strong tradition of using linux for it because of projects like openmosix, and many proprietary scientific software is usually linux-compatible. That desn't mean you can't run it on freebsd, or create MPI applications, or use BSD nodes on your sun gid engine cluster.
Ports are upgradeable at anytime, and freebsd has somewhat frequent releases. If you really need a scheduler for those 48 core, you probably would be better with solaris than with linux. But yes, often server-grade hardware lacks freebsd drivers, and most vendors don't correct defects on drivers not supplied by them. Some vendors don't correct them at all, but it is somewhat easier to fill a problem report when using a supported operating system.
Can't you install freebsd packages via puppet? The upgrade path should be the same. The dependency problem is transversal to most unix so's, including linux (desktopbsd has a workaround for it when using.pbi). I usually update php using make deinstall/reinstall, and have my database instances on another jail. also, usually the php instalations are jailed, so a zero-day exploit might access or wipe the jail data, but won't expose the whole server. If you're running 32 bit VMs, you could probably even install the linux php binaries and run them (though at that point, you may be better served running centos or some other linux distro).
I actually use mostly FreeBSD on servers because the maintainability and time-savings. I've never had a problem with updates not being accessible (cannot say the same for Solaris), and most of my installations work out-of-the-box. They do not have some fancy half-assed X wizard to tune some config file (or some xml that will later be translated to some config file), and requires that the administrator has an idea of what he's doing. The problem is, is cheaper to hire average administrators that can run a linux configuration script or a windows wizard, than one that actually knows what he's doing. And while companies probably don't need a wizard running the servers, a competent unix sysadmin will be competent and productive on freebsd, linux or solaris. The problem with modern operating systems isn't stability anymore, but how they allow the sysadmin to diagnose, troubleshoot and recover software and hardware errors. On that field, both solaris and freebsd are much more advanced than linux or windows.
There is desktop virtualization - VirtualBox. Also, some versions of vmware workstation used to run on it, don't know if it still is an option. For unix-stuff prototyping and development, I find jails far more versatile and fast than some VM running on a desktop. Granted, I don't need X (though you can install it on a jail with some tweaks), but works for me.
Opinions may vary. About a year ago, I benchmarked postgresql on a quad-core xeon, with freebsd 8.1, freebsd-zfs, centos 5.4 and netbsd 5.0. Netbsd performance was less than 1/4 of freebsd, and centos 5.4 was the clear winner. I never run all the tests on netbsd, because the battery of tests that had taken around 12h in freebsd would take about a week on netbsd.
What is the efficiency ratio? 60%? Less? You can recover some energy, but it uses a whole lot more energy than it produces. It works, but is far from being efficient.
Yeah, the power station shuts off the generators when thousands of people turn off standby equipment (not really). Electricity is one of those things that can't be stored efficiently when generated - it is either used or dissipated as heat on the cabling structure. And while the electric grid uses provisioning to peak periods, the provisioning is provided by fast cycle power stations (hydro, nuclear, etc) and not from the most polluting ones (gas, coal, etc). The furnace of an old coal station can take up to a month to heat to the level needed to produce electricity, so it's not difficult to imagine that is better to keep them running just in case than turning them off. In fact, by turning appliances on and off during expected hours, you may contribute to generate more pollution by increasing the load on the grid by increasing load on peaks.
That said, I also live on a european country where electricity isn't cheap,and yes, you can save some pounds/month by actually turning off every standby equipment.
You already have on the market hi-speed PCIe networkable adapters, a form of "external PCIe", at a fraction of the cost of 10G gear. I don't have at hand the manufacturers link, but you can check the available molex connectors (http://www.andovercg.com/datasheets/molex-74546-0813.pdf). These adapters are a big thing, not only for storage and display interfaces, but also for grid computing - you can copy memory data between computers at bus speed.
I guess the parent post was referring to the fact that the news agencies are global news agencies - a photo staged by a photographer in Afghanistan can end up in media all over the world. Many news outlets have correspondants, but even those buy news and content from the global agencies. And yes, that includes France, and many other countries with strict laws about that kind of procedure. There is a somewhat acceptable code of ethics followed by many professionals (http://www.nppa.org/professional_development/business_practices/ethics.html), but at the end of the day, it's like any other business - it's the dramatic results that make money, and not accuracy.
The video card memory isn't mapped onto physical RAM, but onto available linear address space, as other devices. I'm not familiarized with PCIex, but at least on AGP (and previous interfaces), you can specify the aperture window size (the size of the mapped window), that can be a fraction of the available video card memory. On XP32 pre-SP2, you could access the full 4GB of RAM, assuming your hardware and drivers supports it, using PAE (enables 36-bit translation using page sizes of 2Mb or 4Mb). Apparently on XP32 SP2, some PAE funcionality was disabled due to driver problems (http://support.microsoft.com/kb/888137). Also, Windows Server 2003 enterprise (32 bit) uses PAE to access upto 64Gb of physical RAM (SP1).
There are some performance issues regarding the use o PAE due to" windowing" of the available RAM, excess memory allocation, and potential swap file degradation, but a couple of years ago, most laptop manufacturers had available some models with 4GB of RAM, gigabyte graphic card and a 32 bit operating system.
My big problem with browser memory consumption is that usually a usable browser stack (browser+java+flash) is 32 bit, and the 2GB memory limit is a real problem.
Regarding memory management, there's a difference between "system" memory and dynamic(heap) memory - the heap memory is allocated at the start of an application using the operating system API, and can be grown or shrinked if needed.The available heap space is usually managed by a lightweight library (usually part of the compiler runtime), and eventually issues the growing and shrinkage requests. The problem is, since the heap memory manager has no control over virtual addressing (ie cannot remap different blocks to contiguous areas), and usually you have thousands of allocated application blocks with very small sizes, fragmentation is a problem. Even with good memory allocation algorithms (best-fit, automatic block defragmentation, etc), there are a ton of unused blocks between used ones, and within the allocated heap space, so even if the memory manager wanted to return it to the operating system, it can't. If the operating system needs the (paged) memory allocated to a given application, it has no control over the inner workings of the heap memory manager, so it will flush the needed space to the swap file, usually using a glorified LRU implementation. If the application has allocated 1GB of memory, and it has the first 50Mb and the last 50Mb in use, but the rest remains free, you have less 900Mb of memory to be used as system cache/buffers/whatever, and, if needed, will have the impact of the swap in/swap out of the unused pages.
I think Chrome approach of multiple processes (and multiple heaps), one for each window is a much memory friendly approach than Firefox.
Recently I've come across a controller that apparently corrupted SATA disks - a disk connected to a specific channel died 3 times within 2 months. Changed to another channel, and works like a charm, so you may have some problem with your SATA controller.
You could try to use master/slave replication to do a binary backup - sychronize the slave, then shut it down and copy the datafiles. The backup is somewhat version-dependant, but it is a lot quicker than running a sql script. Also, you may want to have a look at cdp tools.
Slashdot Mirror
When someone suggests Dia as an alternative to Visio, I'll laugh. Ahah :) Even Smartdraw, a somewhat easier-to-use-and-a-lot-cheaper alternative to Visio, pales in comparison. It's a bit like suggesting ms paint for cad work.
1) color management and color management support by application software
2) screen real estate - in windows, I get much more workable screen area at the same dpi than any Xorg configuration. I've installed recently windows 8 preview on a XM, booted at 800x600 and no widget, button or whatever was out-of-screen
3) software UI consistency - every windows program looks the same, and the options are at a predictable place; the ribbon UI breaks this, but only until every other program has a ribbon (I hate it)
4) decent clipboard - if I copy something from an application and paste it on another application, I really don't care about what widget toolkit app A is using and if B can decode the format. Vector, image and formatted text copy/paste in xorg are a nightmare
5) stability - windows crashes more than my bsd servers, but every x/xorg I've tried is much more unstable than any windows I've used since NT3.5. On a *nix desktop, having X crashing is almost the same as rebooting
6) decent font support. Windows comes with some nice system fonts, and very good TTF support. Font rendering in windows has some nifty algorithms thay maybe many don't like it, but I do - I can distinguish between semi-bold and bold text
7) explorer performance - try to open a folder with 500 hi-res pictures and wait for the previews in gnome. Now imagine it if it was 2 thousand pictures
8) consistency - in windows, you can navigate trough menus and usually find the options you expect on the place you are looking
9) software variety - you have virtually everything for windows
10) ease of administration - the granularity of the administration knobs (local policy or domain policy) is something almost unheard on *nix destop distros
11) consistent upgrade path - there's a video somewhere of somewone who upgraded windows from windows 1.0 to vista and documented the process. I usually don't upgrade windows installations - I prefer to rotate the disk as a backup and do a clean install - but it is nice to have options.
12) print spooler and print support - windows has ONE print spooler and very good usermode driver print support. That translate to not having to use print processors and interpreters, LPD and CUPS, and subpar generic drivers. There are exceptions - and I even somewhat like CUPS, even if now is an Apple product - but the ppd drivers are usually not that impressive, even when supplied from the manufacturer
13) RDP(v5) - there is no competing technology at the same price. Yes I've tried FreeNX. And VNC. And X tunnedled trough ssh. FreeNX (or whatever was called previously) is quite good, but still no match for RDP
14) multimedia - I don't need Jack funcionality, I need to play multiple sound streams on the same computer without chmod voodoo or to know about audio toolkits. Just need to quickly play a video with my favourite player without needing to stop my mp3 player
15) multiple monitor support. It just works. and you can rotate the screen, or send it to a projector via wireless.
Many/all of these points are addressed by other operating systems, such as MacOS X, BeOS/Haiku (go Haiku!), etc. I've been using windows on desktop since 3.0, so my list is probably biased, because I "hate" x/xorg with passion. Btw, I'm also an unix professional, and use BSD (FreeBSD/OpenBSD) as a part of my daily job, and I've been trying to convince myself to use x/xorg without success since 1995.
Apparently the numbers were only high enough to cause an uproar after the start of persecution of jews in Germany, some year before WWII.
It is easy to have a prosperous nation when you have wealth (from all those fleeding immigrants) and access to unlimited credit. In many regions of the world minorities are slaughtered before they can afford to flee the country they live in, so yes, I'm talking about jewish wealth. If I was one of the minorities being persecuted in Germany and had the choice to leave the country, I'd do it in a hearbeat. But while the jews are no different from other minorities, such as tutsi in Rwanda, the victims of serbian ethnic cleansing, the curds in Iraq (and during Allied supervision of the post gulf war), etc.
Israel allies don't really care for justice or amend of past mistakes - they only care about money. They are the same that overthrown gadaffi -a merciless dictactor, but with a stable country - but in Syria the bodies keep piling and they don't care. If Syria was a major oil and gas producer, they'd sing a different tune. Or maybe if they were jews, they could aspire to have first-class treatment.
Thats what happens when outsiders decide to create a country where some other outsiders live, without regard for the people that actually live there, and their neighbours. If the WWII never happened (and the holocaust), do you believe Israel would exist?
I was referring to the date Britain took control of the territory, aided by the Jewish Legion.
Except that they also use the "self-defense" excuse to attack without military provocation. Can your colleagues punch you in the face now, because someday you may be an asshole to them? Tought so.
Given the military capability of Israel, usually their self-defense is like shooting the kids who threw rocks at your window with a laser-guided missile.
So, because they have been kicked around and harassed, they have the right to do as they please? And how convenient of you to omit the influx of jews to what now is Israel after they assisted Britain in taking control of the territory in 1917. But when the arabic population revolted against what could have been seen as an "invasion", and the brits imposed restrictions on jewish immigration, they turned on them. During the post-WWII era, the palestinian territories were flooded by holocaust survivors.
Now think about this - pick a nationality you don't like very much and/or has different values than yours, imagine that suddenly your whole neighbourhood is populated by those guys, and when you complain they tell you "god gave me this land". You call the authorities and the court decides "ok now your whole neighbourhood is a different nation where their law and values are followed - not yours". Would you accept it?
It is hard to be sympathetic to the Israeli position- they are a nuclear nation, they have as allies some of the more powerful countries in the world, and israeli jewish interests have a lot of weight in international politics - and not necessarily those of the muslim population who also are israeli citizens. Also, the hipocrisy of some Israeli allies is staggering - I'm shure the american right wing would be delighted to have a new nation forced by the UN on their territory, called for example "new new mexico", or "chinazaquistan", as a way to solve their immigration problems.
I'm not defending the extremism against Israel taken by their neighbours (and religiously I really don't care about muslim or jewish deities, or any kind of deities), but one must be blind to not see the manipulation and hipocrisy of their government, and their seemingly overwelming political and economical influence.
I think you are talking about java and eclipse development tools that reportedly run faster using Linux ABI in FreeBSD than in native linux. If I recall correctly, at the time, the mainstream linux kernel was the 2.4 line, that had several design issues that impacted performance on many applications. 2.6 is a whole different beast. The Linux ABI is a nice FreeBSD feature (and currently is 2.6 compatible), but there is no support for 64 bit binaries. Also, some syscalls are not implemented, as well as some exotic functions such as LDT manipulation and virtual 86 mode. The biggest gripe I get with the Linux ABI are the userland environments available on the ports tree - some are old, some are deprecated and/or forbidden. It would be awesome to fire up a jail running a linux userland to use linux-only software, even if limited to 32 bit - Zend Encoder comes to mind.
He can build the package on a separate machine, with the fixed dependencies, and then install it on the VMs via puppet. Package building on FreeBSD isn't as simple as it should be (I prefer the simplicity of the OpenBSD ports in this regard, but I don't recall any broken dependency incident when building PHP. It is not a task I do everyday - once a month or so, but never had any issues as described.
I am a FreebSD user, and most of my work envolves BSD setups.
For most of what I've benchmarked, CentOS 5.4 is faster out-of-the-box than a tuned FreeBSD install. Upon resurce starvation, I find FreeBSD much more stable and easier to recover to a running state. Also, tools like zfs and jails allow me to be able to do more with the machine, and even improve the maintainability of the system. The ports infrastructure is not perfect, but does a good job at it, and I often get less garbage when upgrading packages in freebsd than on linux.
The stability argument is moot nowadays, as I find a CentOS or Debian install as stable as FreeBSD (and forgeting the crappy unstable mess that were the 5.x releases. Most of the arguments against linux or freebsd used in the past arent valid today -as an example, linux has a ton of crappy drivers, freebsd still has a ton of unsupported hardware. Regarding 3D acceleration linux is somewhat far ahead, but behind other operating systems.
I don't play around much with DragonFly, but I think the direction they're heading is grid computing, and not necessarily SMP. The project implements some interesting concepts, and are one of the few operating systems actually trying something "new", and for that they deserve all the merit they can get.
AFAIK most of the top 500 supercomputers run a heavily modified and customized version of linux, not your run-to-the-mill distro. Given that most BSDs were late to the "SMP race" and that Linux already supports a set of architectures at least as vast as netbsd, it is the ideal candidate to be tewaked and installed on those supercomputers. If by supercomputers you ment clusters, yes, there is a strong tradition of using linux for it because of projects like openmosix, and many proprietary scientific software is usually linux-compatible. That desn't mean you can't run it on freebsd, or create MPI applications, or use BSD nodes on your sun gid engine cluster.
Ports are upgradeable at anytime, and freebsd has somewhat frequent releases. If you really need a scheduler for those 48 core, you probably would be better with solaris than with linux. But yes, often server-grade hardware lacks freebsd drivers, and most vendors don't correct defects on drivers not supplied by them. Some vendors don't correct them at all, but it is somewhat easier to fill a problem report when using a supported operating system.
Can't you install freebsd packages via puppet? The upgrade path should be the same. The dependency problem is transversal to most unix so's, including linux (desktopbsd has a workaround for it when using .pbi). I usually update php using make deinstall/reinstall, and have my database instances on another jail. also, usually the php instalations are jailed, so a zero-day exploit might access or wipe the jail data, but won't expose the whole server. If you're running 32 bit VMs, you could probably even install the linux php binaries and run them (though at that point, you may be better served running centos or some other linux distro).
I actually use mostly FreeBSD on servers because the maintainability and time-savings. I've never had a problem with updates not being accessible (cannot say the same for Solaris), and most of my installations work out-of-the-box. They do not have some fancy half-assed X wizard to tune some config file (or some xml that will later be translated to some config file), and requires that the administrator has an idea of what he's doing. The problem is, is cheaper to hire average administrators that can run a linux configuration script or a windows wizard, than one that actually knows what he's doing. And while companies probably don't need a wizard running the servers, a competent unix sysadmin will be competent and productive on freebsd, linux or solaris. The problem with modern operating systems isn't stability anymore, but how they allow the sysadmin to diagnose, troubleshoot and recover software and hardware errors. On that field, both solaris and freebsd are much more advanced than linux or windows.
There is desktop virtualization - VirtualBox. Also, some versions of vmware workstation used to run on it, don't know if it still is an option. For unix-stuff prototyping and development, I find jails far more versatile and fast than some VM running on a desktop. Granted, I don't need X (though you can install it on a jail with some tweaks), but works for me.
Opinions may vary. About a year ago, I benchmarked postgresql on a quad-core xeon, with freebsd 8.1, freebsd-zfs, centos 5.4 and netbsd 5.0. Netbsd performance was less than 1/4 of freebsd, and centos 5.4 was the clear winner. I never run all the tests on netbsd, because the battery of tests that had taken around 12h in freebsd would take about a week on netbsd.
What is the efficiency ratio? 60%? Less? You can recover some energy, but it uses a whole lot more energy than it produces. It works, but is far from being efficient.
Yeah, the power station shuts off the generators when thousands of people turn off standby equipment (not really). Electricity is one of those things that can't be stored efficiently when generated - it is either used or dissipated as heat on the cabling structure. And while the electric grid uses provisioning to peak periods, the provisioning is provided by fast cycle power stations (hydro, nuclear, etc) and not from the most polluting ones (gas, coal, etc). The furnace of an old coal station can take up to a month to heat to the level needed to produce electricity, so it's not difficult to imagine that is better to keep them running just in case than turning them off. In fact, by turning appliances on and off during expected hours, you may contribute to generate more pollution by increasing the load on the grid by increasing load on peaks.
That said, I also live on a european country where electricity isn't cheap,and yes, you can save some pounds/month by actually turning off every standby equipment.
You already have on the market hi-speed PCIe networkable adapters, a form of "external PCIe", at a fraction of the cost of 10G gear. I don't have at hand the manufacturers link, but you can check the available molex connectors (http://www.andovercg.com/datasheets/molex-74546-0813.pdf). These adapters are a big thing, not only for storage and display interfaces, but also for grid computing - you can copy memory data between computers at bus speed.
And Carla Bruni (staged) nude photos. Not that I'm complaining or somethin'...
I guess the parent post was referring to the fact that the news agencies are global news agencies - a photo staged by a photographer in Afghanistan can end up in media all over the world. Many news outlets have correspondants, but even those buy news and content from the global agencies. And yes, that includes France, and many other countries with strict laws about that kind of procedure. There is a somewhat acceptable code of ethics followed by many professionals (http://www.nppa.org/professional_development/business_practices/ethics.html), but at the end of the day, it's like any other business - it's the dramatic results that make money, and not accuracy.
The video card memory isn't mapped onto physical RAM, but onto available linear address space, as other devices. I'm not familiarized with PCIex, but at least on AGP (and previous interfaces), you can specify the aperture window size (the size of the mapped window), that can be a fraction of the available video card memory. On XP32 pre-SP2, you could access the full 4GB of RAM, assuming your hardware and drivers supports it, using PAE (enables 36-bit translation using page sizes of 2Mb or 4Mb). Apparently on XP32 SP2, some PAE funcionality was disabled due to driver problems (http://support.microsoft.com/kb/888137). Also, Windows Server 2003 enterprise (32 bit) uses PAE to access upto 64Gb of physical RAM (SP1).
There are some performance issues regarding the use o PAE due to" windowing" of the available RAM, excess memory allocation, and potential swap file degradation, but a couple of years ago, most laptop manufacturers had available some models with 4GB of RAM, gigabyte graphic card and a 32 bit operating system.
My big problem with browser memory consumption is that usually a usable browser stack (browser+java+flash) is 32 bit, and the 2GB memory limit is a real problem.
Regarding memory management, there's a difference between "system" memory and dynamic(heap) memory - the heap memory is allocated at the start of an application using the operating system API, and can be grown or shrinked if needed.The available heap space is usually managed by a lightweight library (usually part of the compiler runtime), and eventually issues the growing and shrinkage requests. The problem is, since the heap memory manager has no control over virtual addressing (ie cannot remap different blocks to contiguous areas), and usually you have thousands of allocated application blocks with very small sizes, fragmentation is a problem. Even with good memory allocation algorithms (best-fit, automatic block defragmentation, etc), there are a ton of unused blocks between used ones, and within the allocated heap space, so even if the memory manager wanted to return it to the operating system, it can't. If the operating system needs the (paged) memory allocated to a given application, it has no control over the inner workings of the heap memory manager, so it will flush the needed space to the swap file, usually using a glorified LRU implementation. If the application has allocated 1GB of memory, and it has the first 50Mb and the last 50Mb in use, but the rest remains free, you have less 900Mb of memory to be used as system cache/buffers/whatever, and, if needed, will have the impact of the swap in/swap out of the unused pages.
I think Chrome approach of multiple processes (and multiple heaps), one for each window is a much memory friendly approach than Firefox.
Recently I've come across a controller that apparently corrupted SATA disks - a disk connected to a specific channel died 3 times within 2 months. Changed to another channel, and works like a charm, so you may have some problem with your SATA controller.
You could try to use master/slave replication to do a binary backup - sychronize the slave, then shut it down and copy the datafiles. The backup is somewhat version-dependant, but it is a lot quicker than running a sql script. Also, you may want to have a look at cdp tools.