I'd like an OS to let me specify the 'world' my application runs in (which processors, how many, etc.) These interfaces are available in Windows at run time (the task manager will let you adjust where a running task can go).'" Isn't this *exactly* what numactl gives you (hint, you don't have to run numa for numactly to allow processes to be bound to specific sets of processors).
The point is php did the work for him, that php was faster, and that all the failings with the php codebase were a matter of inexperience not of the language itself (i.e. his last point, though it could have been stated more plainly). A new programming tool will more often than not feel more impressive because it forces you to start from the beginning again, and you do things more easily and simply with successive iterations, even if you forced yourself to do so with the same language.
With Rails, the problem is that the whole point, the shortcuts that the framework provided, didn't map to what he wanted to do. Your response is that 'well, he could have used Ruby/Rails, and opted not to use any of the shortcuts', but that robs Ruby/Rails of the entire advantage. Sure it's complete and you can code whatever you want, but if the shortcuts don't map to what you are doing, then it's no better than using plain php. Frameworks with very high-level primitives typically require that you shoehorn your project into their model more than the likes of php.
Overall, his problem was that he picked a technology for the sake of the hype around it without sound technical reasoning. On the brightside, along the two years of bumbling along trying to follow the hype, he learned what he screwed up his first time around with php, and could make it better.
My thought is don't jump to new technology just because. Be extra wary of things with tons of hype surrounding them, and take the time to evaluate new technology (ensure the people evaluating has used the current tech so they can make a fair comparison). The adage 'if it isn't broke, don't fix it' carries a lot of weight. That said:
Thinking about ditching C for Java? Think again. I will say a lot of companies have successfully ditched C for Java, or at least C for C++. I personally dislike the Java implementations I've seen, Java programs I've used still feel slow and stick out like a sore thumb on my desktop (I know, SWT was supposed to help this, and I do have some SWT apps, they still somehow feel bizarre). My personal opinion is that Java as widely implemented is some weird thing with the worst of all worlds. It requires compilation, but still acheives none of the performance benefits of C. I'd be inclined to use php, perl, or python depending on the scenario if Java was on the plate. But my rant aside, you can't pretend Java has been a total flop.
Thinking about ditching Windows for Linux? Think again. Now that's just flamebait. My house is 100% Linux, my work is 98% Linux, and it's truly a great thing. One problem I would say would be overaggressive marketing to the masses, convincing people to try Linux before removable media was magically handled, before the desktop file managers had decent abstractions for common networking protocols (i.e. SMB), before something like NetworkManager existed to 'deal with your network' for you. With these added, the technical and non-technical advantages of Linux are available combined with the 'ease'-of-use MS offers. Meanwhile, some of the fast-tracks to having all that ease-of-use MS employed left gaping security issues to tackle they still struggle with, as well as a situation where the MS experts faced with a somewhat broken install just reinstalls because they feel its hopeless.
Linux also represents something significant. Essentially it has come to represent a set of software that does things in a standard way that multiple businesses can implement and compete in a compatible way. This is the key thing that pushed the PC architecture, competition. You can choose Ubuntu, Red Hat(Fedore/CentOS), SuSE, Debian, Mandriva, Gentoo, or any number of others, and still run the same software. You can pull one together yourself if so inclined, or use a community maintained one, and/or have commercial support. MS rode the PC platform to success because they were the only software company to support it on arbitrary hardware vendors. So powerful was this advantage that the industry essentially gave them the monopoly before any other company had a serious offering to consider under the same terms (They had it pretty well gotten by Windows 3.0). Now Linux can represent the same phenomenon, but for the operating system. In order to overcome MS, it's had to be truly Free, but there remains a healthy commercial environment around it.
Of course cell phones are not free, most people pay for them over time, usually paying $1.5k over a two year period. Yes, so on this count, the iPhone is the worst of both worlds. It *requires* a two year plan (at least, to be legitimate within the Apple/AT&T vision and have cell service), *and* costs about the same as the unlocked phones (and before costed significantly more), and that's with requiring more expensive data plans. I wasn't surprised to see the iPhone price, but I was surprised to see them declare that there would be no contract-signing subsidizing of the cost, but you'll have to sign it anyway.
It would be interesting to know the details of the Apple/AT&T deal, if AT&T pays some flat amount of money for the privilege, or pays at some scale with respect to Apple phone sales, or if somehow Apple figured they needed to partner with one and only one cell service provider to make a successful launch, and it is even financial-wise.
I think I've seen entirely too much footage absolutely convincing me that with the deployment of Tazers, that some of the police use it too readily. I remember seeing footage where a guy was asked to turn around, complied, asked to put his hands behind his back, he did. They cuffed one arm and couldn't quite reach the other wrist, and in trying to pull the other arm closer, they make some comment about 'stop resisting, just let us cuff', and then within five seconds of no obvious struggling, they taze. The worst he could have been doing was holding his arm stiff, but he wasn't actively doing anything and had appeared to be very compliant to that point.
I couldn't watch the video with sound up, so I don't know when the Tazing occurred, but it is safe to say people have been Tazed by police with much less justification than this guy.
First off, you are right that the summary at least misunderstands SMP. Technically speaking though, AMD doesn't *technically* become NUMA until multi-socket comes into play, so it's fair to call a single-socket AMD system a SMP system. So for most home systems, AMD would be SMP if dual core, tri-core, quad-core, etc.
However, you could turn any SMP system into ASMP. ASMP can also refer to how the OS uses the processors (i.e. if one processor services and only services interrupts, it's ASMP). You could set the interrupt mask to point to all one processor, and use numactl to bind every process on the system to not hit that CPU. This isn't overly usefull, but you could technically claim ASMP then..
You can do 4 objects and connect them all without oven using another layer. Picture a triangle with the other component in the middle. Connect every vertex to the middle. Make the traces to the middle zigzag a bit to even out the trace lengths, and boom, fully connected without any intersections. Not saying this is how things are done, mind you, but it is a silly argument to say three cores are good because they can be connected trivially. 3-core cpus are all about yield. Being able to sell components that had a flaw in a core, without reverting all the way down to a two core part (and by extension the two core price point), is important.
All that said, SMP has nothing to do with an even number of processors/cores. It just means each processing element of a system is roughly equivalent. So you have a choice of three parts to schedule something on, the scheduler can know all three are equally capable and the heuristics for processor selection are straightforward. ASMP typically has specific roles for each part (i.e. a dedicated processor for interrupts, etc etc)
I would say that if you managed to slip in peacefully and didn't deprive anyone of their seat, conduct yourself peacefully, don't push the stadium over its maximum occupancy, etc etc, then yes, it is truly harmless. Hard to gauge the maximum occupancy, especcially not knowing how many other people hyptohetically sneaked in, so it would be hard to know ahead of time if you are creating a dangerous situation, so it's not a good idea, but the act isn't intrinsically harmful. Ethically justified, hard to say, but not impactful to anyone signifiantly. It's not like you get anything other than to listen and interact with other fans. If anything, helping fill the house enriches the experience in a NIN concert.
I really wonder how much difference ubiquitous solar power would make compared to our typical activities of clearcutting/paving/etc etc. If anything, I would think hypothetically high efficiency solar panels would serve to offset what we generally do (i.e. dark colored roofs and pavement are the norm, replacing lighter colored plantlife). But I don't really know how all that goes.
Alternatively, set up extra-planetary solar collectors not in our path of sunlight. Of course, then you have to somehow transfer that energy..
For example serial. What other way can you get textual BIOS output on a server for logging and such (for that matter, grub output and early kernel output). Granted, the desktop world is different, but serial has no replacement in at least one role.
I personally don't know if any role that floppies, IDE, and parallel ports fulfill that aren't technically served by a better technology (add PS/2 ports if you wish), but at the same time, I haven't bothered to buy a new printer in many years, and as such, still parallel port. Same with an old scanner. Sometimes technology is 'good as I'll ever need' in a particular genre, and I don't feel bothered to upgrade. And even if you expect upgrades to be 100%, there is a time during which you'll need both in one place for things like hard drive data transfer.
The core question here is what is the cost of adding the IDE ports, floppy ports, serial, parallel, etc. Well, until very very recent chipsets, the system chipset would have integrated IDE controllers, so the cost-add is only that of passive traces/connectors being put down, which is extremely small. Even with chipsets that have sunsetted integrated IDE, the cost of IDE chips is low. The components to drive a serial/parallel port are similarly either integrated or unbelievably cheap to add. The short of all this is that in the x86 motherboard space, it is a market that demands every corner be cut as much as possible. You can be assured that the cost delta of supporting legacy has been examined time and time again, and it's just not that significant a difference. Now compare this to putting all that only on adapter cards/usb. A PCI adapter can't hook into the integrated chipset, so it has to have active components. An adapter faces a more rigorous interoperability requirement, therefore needs more testing of that nature and potentially more development. Ultimately, the combination of board+card to achieve what a competitor does all-in-one will end up being more expensive *and* less convenient to a customer needing it. Especially since in most cases that you advocate, the board manufacturers would replace the cost of a simple passive connector add with a card/active chip/passive connector add. You could say the chip manufacturer's would be the ones to drop support, but the justification is just not that good and won't lead to significant cost savings (at least, for the n-1 case, supporting too many generatinos of old tech would get more complex in a non-linear way).
Actually, most people don't have copyrights over the material that gets pulled off. Actually, most do. Most may not realize it, but copyright is pretty much automatic. You don't even need to post a copyright notice of any sort or register it with anyone, so long as you can provide reasonable proof of authorship. Of course, you can pay for a solid defense in an official way..
But seriously, a good deal of the health of the Linux community has been due to the GPL. It has also been the reason why companies are so fearful, yes, but once in, they generally end up doing the right thing because of the licensing terms.
There are no shortage of commercial products with their roots in a BSD. The problem is they most often don't bother to contribute work back. There is some mindshare that letting upstream maintain non-specific stuff for you is inherently better, but at the same time it takes effort to decide where that boundary is, and many companies don't bother. The BSD projects that have code contributed by companies back are largely in the context of using them under Linux, and as such they are already in the habit of doing that.
Is the difference between sleep and off obvious to a non-technical person? Is there a way to turn it off really? Does the iPhone explicitly show that it is in sleep vs. off? While in sleep, how is a typical user supposed to know it's still transmitting? Can you point to any other cell phone that behaves this way?
My laptop wireless interfaces are not active during sleep. The closest analogy I can think of is the Wii, and that isn't a mobile device attempting to use a quite likely charge-per-usage network constantly.
Apple's standard is simple and straightforward. There is a balance here to be struck between convenience and, well, this. At the least, the standby/sleep mode should have a fee-less way of determining if the transaction would be built into the plan before executing it. If it would cost extra, and you are in sleep mode, don't do it. Even then that won't be convenient for the common user to comply with airline requests, with most people thinking it would be the one button that makes it go dark.
Opteron at least on floating point is lower than Woodcrest/Clovertown IPC in 64-bit. Note the top500 and the increase of Intel presence as of the Core2 generation. Barcelona is supposed to either meet or beat the Intel floating point IPC, but that's yet to be proven publicly. There is at least one significant 64-bit operation that Core2 creams AMD with. I don't know much about other types of instructions in general though.
I agree though, AMD's architecture scales *much* better with socket count and memory architecture wise blows Intel away.
If the patents around WAFL/ZFS went away, a kernel-level ZFS/WAFL-like implementation would be possible without concerns over the patents. However, my suspicion is that the two parties would ultimately settle out-of-court before risking such a thing. If it went to court, they would be pretty confident that they wll either win, or at least prove unclean hands such that it mostly ends up still being even, but still allowing them to sue third parties. (BTW, IANAL)
No, there is no DTrace in Linux, and nothing close. People who truly grok the utilities historically considered 'development tools' achieve the same things. Attaching to running processes to monitor system calls being made, profiling, monitoring file handles, and attaching debug on the fly. Every time I've had DTrace demoed to me, they focus on 'look, I didn't start under a debugger or anything and 'poof', I'm watching file-handles as they open and close!'. Maybe all the people coming to demo DTrace are missing the point, but generally they aren't aware that the likes of strace and gdb can attach to a random running process and follow it. Even without debug symbols there is a lot the 'debug' tools can do. Sun is just marketing those facilities to admins (of which I'd say >90% never groked the concept of 'development' tools and how they could be useful) and providing a more coherent interface to it.
Containers, though, give you away. Which other platform allows you to, like, start 1000 virtual machines, each consuming 2-3 MB and virtually no CPU resources as long as you don't run anything in there ? Not Solaris. The whole *point* of containers is explicitly *not* to have virtual machines. You have, essentially, uber-chroot, with nicely partitioned procces space and userland storage. You run everything under the same kernel image, the kernel is just smart about having really independent environments. Sun likes to boast about 'BrandZ', and some get the impression they are running, for example, a linux system, but it's really just a container with linux libraries leveraging the Solaris kernel linux compatibility. Virtuozzo (OpenVZ) would be the linux side of all this.
There is a lack of hardware support, no doubt. But also, the rumoured lack of stability is not trustworthy. Except, you run crappy, faulty, hardware. That is simply a poor defense. It's a manifestation of a platform trying to open up to a wide audience and as such facing variety those drivers have *never* been scrutinized to accommodate. This is a normal process for a project facing a broader audience, and it's just not reasonable to deny the fact that drivers that have yet to be tested on all the hardware it tries to support will be subject to problems.
I will again concur on the Nexenta point, but it's more of a 'not invented here' syndrome. Their trumpeting project Indiana so much is sufficient proof that they want to hop on that train, they just don't want to be obviously following someone else's lead.
Sun has done an excellent job of astroturfing. I know a lot of technical people who have tried it once again, and got the 'neat' factor of ZFS, was not that impressed with DTrace (we know how to do most of this sort of stuff in linux already), and containers, well, are nothing unique to the platform. So ZFS remains the cool thing that, while Linux has facilities to kinda-sorta get there, can't get there as smoothly and flexibly. Meanwhile, they were bitten by a distinct lack of drivers, and their random whitebox platform they used to evaluate was being strangely flaky in the face of Solaris when it seemed solid with Linux.
So on the technical front, there remain kinks to work out. In the meantime, Linux has incredible momentum, incredible talent in the market, and from a business standpoint, is in an advantageous position. Linux has more corporate backing (you want serious software support for Solaris, you have only Sun to choose really, while in Linux, well, at least Novell and RedHat are serious software support contenders, and more hardware vendors embrace Linux than Solaris).
The other sad thing was the Solaris platform package management. Nexenta was a refreshing thing to evaluate, but looking at the community at large it seems Nexenta gets the shaft. It's all up to Indiana to see if they can pull off a well-accepted, decent package/repository system. I have to admit, this is by *far* the biggest thing Linux platforms have going for it (apt/yum) and very much outweighs the benefits of ZFS (it's like apples and oranges, true, but when you have to pick one or the other...). Of course, the Nexenta situation points to them not pursuing the other thing they need to be a Linux contender, they'd have to allow other companies to have control and be able to provide software support on their own without any help or money exchange with Sun themselves. The question is if they did that, would Sun's share of the Solaris market still be more than the current Solaris market in the face of a dominant Linux market, and I really have no idea. They might just have to lose out on Solaris to make it have a chance, and that really gets them nowhere. It's a fine line to walk and it wil be interesting to see what they do to try to pull it off.
I can do XFS too (I know you made a mistake, and mean ZFS). However, I will point out: $df -h . Filesystem Size Used Avail Use% Mounted on/mnt/t/something 16T 1.1M 16T 1%/mnt/t/t $df -k . Filesystem 1K-blocks Used Available Use% Mounted on/mnt/t/something 17100669952 1056 17100668896 1%/mnt/t/t
I just ran this on my laptop (an 'average' system, though I assume your system with 16 TB of storage is not really 'average'. I too can have big block devices with a single filesystem, big deal. Go commercial, ala GPFS and you can do bigger (http://en.wikipedia.org/wiki/Comparison_of_file_s ystems). I just have a hard time having enough storage to build such a filesystem. The biggest real block device (not sparse) I have readily available not on GPFS is an 8 TB ext3 filesystem.
ZFS's power is not the filesystem size. It unifies a lot of things historically in different layers. I.e. software raid, storage pools, dynamic new filesystems, long term snapshotting. Most of these can be done without ZFS, but the creating filesystems and long-term snapshotting can be done with such ease and efficiency when all the 'layers' work together, and that is what ZFS brings to the table. I will say ext3cow would give me the single feature that most appeals to me about ZFS, and the rest I can do using LVM and such.
In the end, ZFS is the single point that tempts me in general about Solaris, but I'm not about to jump platforms when I know enough 'tricks' to get 'good enough' out of my existing platform.
At least in linux, and I presume FreeBSD's swap strategy is similar, you miss the point. Let's look at two scenarios, one with proactive swapping, one without, and a malloc comes in that exceeds system memory.
Non-proactive case: -kernel sees malloc, knows it lacks physical memory to accommodate, malloc is blocked while kernel does housekeeping. -kernel picks the appropriate amount of pages to write to swap, then writes those pages to swap space, taking a while since block storage IO is excruciatingly slow. -After the extremely long previous step, the memory is freed -the malloc is allowed to continue, after a number of milliseconds have passed to execute the drive write, aside from the drive write, everything was in the microsecond scale, so it was delayed by a factor of thousands.
Proactive case: -The system has some idle time, with nothing immediately better to do, the kernel notes free swap space and flags some appropriate memory as what would be swapped out if and when the system was in need, and copies it to disk, but it *leaves it in memory*. The kernel remembers that while these pages are indeed in memory, it can zap them and be able to restore. This is the critical point, the data in memory has not been *moved* to swap, it has been *copied* to swap. -Program using that data randomly kicks back to life. It's needed data is on disk, but it is a moot point because it is in physical memory too, so it isn't slowed down. The kernel might take this opportunity to re-evaluate things when idle in terms of what it thinks is mostly unwanted pages. -Later on, a program needs to malloc and physical memory is exhausted, the kernel blocks to do housekeeping, finds pages that it knows it has copied to disc, frees them and uses them to satisfy the malloc, within microseconds.
Proactive swapping causes extra IO activity during idle, but does not, if implemented correctly, impact things proactively swapped unnecessarily negatively, and allows swap on actual demand to be nearly trivially fast. It may be wasteful to have gobs of swap, and certainly if the swap has the sole copy of tons of data then performance is hopeless, but don't think seeing the swap used count go up 'mysteriously' without significant mallocs going on that it will impact access to the data written to swap later on.
The point is, the code from inception was organized by the developer as 'either GPL or BSD'. Anyone contributing code to that should recognize that. If someone choses to follow GPL, devs shouldn't get offended, they knew what they were getting into. If you don't want to play by the rules of a segment of a project, go away. Claiming that people cannot strip the BSD license on a redistribution is like saying because one of the licenses is GPL, you must always ship the source code. It's ok to dual-license something, but you have to recognize that each license explicitly grants rights that the other may preclude (i.e. BSD explicitly grants the right to rip code and use as you see fit, without source, which the GPL explicitly forbids). If a project choses to use the code under one license or the other, you simply shouldn't take offense. If you didn't want the code to be distributed under GPL, you shouldn't say it's ok in the first place. If it is more of a 'you must follow the rules of both', it kinda blows up and can't be used in either world. BSD as a whole is setup such that a third party should be able to use it with impunity, and code which may have the GPL apply precludes that. The same people bitching about BSD license being stripped would be happy to see the code redistributed as a binary.
Modern 'supercomputers' are not fundamentally so different from any other network of more than one server. Most run at least one independent kernel instance on each server, and that kernel is generally the same kernel as anyone else would use of that platform. The brains about sending messages efficiently between nodes is generally handled by userspace software (though you might make the claim that low-latency interconnects such as Infiniband and Myrinet are mostly cluster oriented, and have drivers that provide a low level medium for the userspace tools through their kernel drivers, but the story configuration involved none of that). To make something that no one could argue against having the same basic architecture as a 'real' cluster, you just need two-three computers with ethernet connectivity, and download something like OpenMPI and off you go.
Their number sounds about right for 4 2.0 ghz Opterons. If they used core2, they are doing pretty terrible at Rmax/Rpeak ratio. Maybe they have terribly low amounts of ram per core (which precludes tho Top500 benchmark from doing well).
Slashdot Mirror
That means... I really don't want to fall asleep around other men at work?
The point is php did the work for him, that php was faster, and that all the failings with the php codebase were a matter of inexperience not of the language itself (i.e. his last point, though it could have been stated more plainly). A new programming tool will more often than not feel more impressive because it forces you to start from the beginning again, and you do things more easily and simply with successive iterations, even if you forced yourself to do so with the same language.
With Rails, the problem is that the whole point, the shortcuts that the framework provided, didn't map to what he wanted to do. Your response is that 'well, he could have used Ruby/Rails, and opted not to use any of the shortcuts', but that robs Ruby/Rails of the entire advantage. Sure it's complete and you can code whatever you want, but if the shortcuts don't map to what you are doing, then it's no better than using plain php. Frameworks with very high-level primitives typically require that you shoehorn your project into their model more than the likes of php.
Overall, his problem was that he picked a technology for the sake of the hype around it without sound technical reasoning. On the brightside, along the two years of bumbling along trying to follow the hype, he learned what he screwed up his first time around with php, and could make it better.
Linux also represents something significant. Essentially it has come to represent a set of software that does things in a standard way that multiple businesses can implement and compete in a compatible way. This is the key thing that pushed the PC architecture, competition. You can choose Ubuntu, Red Hat(Fedore/CentOS), SuSE, Debian, Mandriva, Gentoo, or any number of others, and still run the same software. You can pull one together yourself if so inclined, or use a community maintained one, and/or have commercial support. MS rode the PC platform to success because they were the only software company to support it on arbitrary hardware vendors. So powerful was this advantage that the industry essentially gave them the monopoly before any other company had a serious offering to consider under the same terms (They had it pretty well gotten by Windows 3.0). Now Linux can represent the same phenomenon, but for the operating system. In order to overcome MS, it's had to be truly Free, but there remains a healthy commercial environment around it.
It would be interesting to know the details of the Apple/AT&T deal, if AT&T pays some flat amount of money for the privilege, or pays at some scale with respect to Apple phone sales, or if somehow Apple figured they needed to partner with one and only one cell service provider to make a successful launch, and it is even financial-wise.
I'll wager a broken fiber in a cable would manifest itself as 'USB2 only' connection.
I think I've seen entirely too much footage absolutely convincing me that with the deployment of Tazers, that some of the police use it too readily. I remember seeing footage where a guy was asked to turn around, complied, asked to put his hands behind his back, he did. They cuffed one arm and couldn't quite reach the other wrist, and in trying to pull the other arm closer, they make some comment about 'stop resisting, just let us cuff', and then within five seconds of no obvious struggling, they taze. The worst he could have been doing was holding his arm stiff, but he wasn't actively doing anything and had appeared to be very compliant to that point.
I couldn't watch the video with sound up, so I don't know when the Tazing occurred, but it is safe to say people have been Tazed by police with much less justification than this guy.
First off, you are right that the summary at least misunderstands SMP. Technically speaking though, AMD doesn't *technically* become NUMA until multi-socket comes into play, so it's fair to call a single-socket AMD system a SMP system. So for most home systems, AMD would be SMP if dual core, tri-core, quad-core, etc.
However, you could turn any SMP system into ASMP. ASMP can also refer to how the OS uses the processors (i.e. if one processor services and only services interrupts, it's ASMP). You could set the interrupt mask to point to all one processor, and use numactl to bind every process on the system to not hit that CPU. This isn't overly usefull, but you could technically claim ASMP then..
You can do 4 objects and connect them all without oven using another layer. Picture a triangle with the other component in the middle. Connect every vertex to the middle. Make the traces to the middle zigzag a bit to even out the trace lengths, and boom, fully connected without any intersections. Not saying this is how things are done, mind you, but it is a silly argument to say three cores are good because they can be connected trivially. 3-core cpus are all about yield. Being able to sell components that had a flaw in a core, without reverting all the way down to a two core part (and by extension the two core price point), is important.
All that said, SMP has nothing to do with an even number of processors/cores. It just means each processing element of a system is roughly equivalent. So you have a choice of three parts to schedule something on, the scheduler can know all three are equally capable and the heuristics for processor selection are straightforward. ASMP typically has specific roles for each part (i.e. a dedicated processor for interrupts, etc etc)
I would say that if you managed to slip in peacefully and didn't deprive anyone of their seat, conduct yourself peacefully, don't push the stadium over its maximum occupancy, etc etc, then yes, it is truly harmless. Hard to gauge the maximum occupancy, especcially not knowing how many other people hyptohetically sneaked in, so it would be hard to know ahead of time if you are creating a dangerous situation, so it's not a good idea, but the act isn't intrinsically harmful. Ethically justified, hard to say, but not impactful to anyone signifiantly. It's not like you get anything other than to listen and interact with other fans. If anything, helping fill the house enriches the experience in a NIN concert.
I really wonder how much difference ubiquitous solar power would make compared to our typical activities of clearcutting/paving/etc etc. If anything, I would think hypothetically high efficiency solar panels would serve to offset what we generally do (i.e. dark colored roofs and pavement are the norm, replacing lighter colored plantlife). But I don't really know how all that goes.
Alternatively, set up extra-planetary solar collectors not in our path of sunlight. Of course, then you have to somehow transfer that energy..
For example serial. What other way can you get textual BIOS output on a server for logging and such (for that matter, grub output and early kernel output). Granted, the desktop world is different, but serial has no replacement in at least one role.
I personally don't know if any role that floppies, IDE, and parallel ports fulfill that aren't technically served by a better technology (add PS/2 ports if you wish), but at the same time, I haven't bothered to buy a new printer in many years, and as such, still parallel port. Same with an old scanner. Sometimes technology is 'good as I'll ever need' in a particular genre, and I don't feel bothered to upgrade. And even if you expect upgrades to be 100%, there is a time during which you'll need both in one place for things like hard drive data transfer.
The core question here is what is the cost of adding the IDE ports, floppy ports, serial, parallel, etc. Well, until very very recent chipsets, the system chipset would have integrated IDE controllers, so the cost-add is only that of passive traces/connectors being put down, which is extremely small. Even with chipsets that have sunsetted integrated IDE, the cost of IDE chips is low. The components to drive a serial/parallel port are similarly either integrated or unbelievably cheap to add. The short of all this is that in the x86 motherboard space, it is a market that demands every corner be cut as much as possible. You can be assured that the cost delta of supporting legacy has been examined time and time again, and it's just not that significant a difference. Now compare this to putting all that only on adapter cards/usb. A PCI adapter can't hook into the integrated chipset, so it has to have active components. An adapter faces a more rigorous interoperability requirement, therefore needs more testing of that nature and potentially more development. Ultimately, the combination of board+card to achieve what a competitor does all-in-one will end up being more expensive *and* less convenient to a customer needing it. Especially since in most cases that you advocate, the board manufacturers would replace the cost of a simple passive connector add with a card/active chip/passive connector add. You could say the chip manufacturer's would be the ones to drop support, but the justification is just not that good and won't lead to significant cost savings (at least, for the n-1 case, supporting too many generatinos of old tech would get more complex in a non-linear way).
Until Netcraft confirms it...
But seriously, a good deal of the health of the Linux community has been due to the GPL. It has also been the reason why companies are so fearful, yes, but once in, they generally end up doing the right thing because of the licensing terms.
There are no shortage of commercial products with their roots in a BSD. The problem is they most often don't bother to contribute work back. There is some mindshare that letting upstream maintain non-specific stuff for you is inherently better, but at the same time it takes effort to decide where that boundary is, and many companies don't bother. The BSD projects that have code contributed by companies back are largely in the context of using them under Linux, and as such they are already in the habit of doing that.
Is the difference between sleep and off obvious to a non-technical person? Is there a way to turn it off really? Does the iPhone explicitly show that it is in sleep vs. off? While in sleep, how is a typical user supposed to know it's still transmitting? Can you point to any other cell phone that behaves this way?
My laptop wireless interfaces are not active during sleep. The closest analogy I can think of is the Wii, and that isn't a mobile device attempting to use a quite likely charge-per-usage network constantly.
Apple's standard is simple and straightforward. There is a balance here to be struck between convenience and, well, this. At the least, the standby/sleep mode should have a fee-less way of determining if the transaction would be built into the plan before executing it. If it would cost extra, and you are in sleep mode, don't do it. Even then that won't be convenient for the common user to comply with airline requests, with most people thinking it would be the one button that makes it go dark.
It has a 'airplane mode' setting. Of course, it certainly isn't obvious to a normal person that an 'off' device could be expected to transmit.
Opteron at least on floating point is lower than Woodcrest/Clovertown IPC in 64-bit. Note the top500 and the increase of Intel presence as of the Core2 generation. Barcelona is supposed to either meet or beat the Intel floating point IPC, but that's yet to be proven publicly. There is at least one significant 64-bit operation that Core2 creams AMD with. I don't know much about other types of instructions in general though.
I agree though, AMD's architecture scales *much* better with socket count and memory architecture wise blows Intel away.
If the patents around WAFL/ZFS went away, a kernel-level ZFS/WAFL-like implementation would be possible without concerns over the patents. However, my suspicion is that the two parties would ultimately settle out-of-court before risking such a thing. If it went to court, they would be pretty confident that they wll either win, or at least prove unclean hands such that it mostly ends up still being even, but still allowing them to sue third parties. (BTW, IANAL)
I will again concur on the Nexenta point, but it's more of a 'not invented here' syndrome. Their trumpeting project Indiana so much is sufficient proof that they want to hop on that train, they just don't want to be obviously following someone else's lead.
A microphone. VOIP with WIFI is kinda useless without a mic...
Sun has done an excellent job of astroturfing. I know a lot of technical people who have tried it once again, and got the 'neat' factor of ZFS, was not that impressed with DTrace (we know how to do most of this sort of stuff in linux already), and containers, well, are nothing unique to the platform. So ZFS remains the cool thing that, while Linux has facilities to kinda-sorta get there, can't get there as smoothly and flexibly. Meanwhile, they were bitten by a distinct lack of drivers, and their random whitebox platform they used to evaluate was being strangely flaky in the face of Solaris when it seemed solid with Linux.
So on the technical front, there remain kinks to work out. In the meantime, Linux has incredible momentum, incredible talent in the market, and from a business standpoint, is in an advantageous position. Linux has more corporate backing (you want serious software support for Solaris, you have only Sun to choose really, while in Linux, well, at least Novell and RedHat are serious software support contenders, and more hardware vendors embrace Linux than Solaris).
The other sad thing was the Solaris platform package management. Nexenta was a refreshing thing to evaluate, but looking at the community at large it seems Nexenta gets the shaft. It's all up to Indiana to see if they can pull off a well-accepted, decent package/repository system. I have to admit, this is by *far* the biggest thing Linux platforms have going for it (apt/yum) and very much outweighs the benefits of ZFS (it's like apples and oranges, true, but when you have to pick one or the other...). Of course, the Nexenta situation points to them not pursuing the other thing they need to be a Linux contender, they'd have to allow other companies to have control and be able to provide software support on their own without any help or money exchange with Sun themselves. The question is if they did that, would Sun's share of the Solaris market still be more than the current Solaris market in the face of a dominant Linux market, and I really have no idea. They might just have to lose out on Solaris to make it have a chance, and that really gets them nowhere. It's a fine line to walk and it wil be interesting to see what they do to try to pull it off.
I can do XFS too (I know you made a mistake, and mean ZFS). However, I will point out: /mnt/t/something 16T 1.1M 16T 1% /mnt/t/t /mnt/t/something 17100669952 1056 17100668896 1% /mnt/t/t
s ystems). I just have a hard time having enough storage to build such a filesystem. The biggest real block device (not sparse) I have readily available not on GPFS is an 8 TB ext3 filesystem.
$df -h .
Filesystem Size Used Avail Use% Mounted on
$df -k .
Filesystem 1K-blocks Used Available Use% Mounted on
I just ran this on my laptop (an 'average' system, though I assume your system with 16 TB of storage is not really 'average'. I too can have big block devices with a single filesystem, big deal. Go commercial, ala GPFS and you can do bigger (http://en.wikipedia.org/wiki/Comparison_of_file_
ZFS's power is not the filesystem size. It unifies a lot of things historically in different layers. I.e. software raid, storage pools, dynamic new filesystems, long term snapshotting. Most of these can be done without ZFS, but the creating filesystems and long-term snapshotting can be done with such ease and efficiency when all the 'layers' work together, and that is what ZFS brings to the table. I will say ext3cow would give me the single feature that most appeals to me about ZFS, and the rest I can do using LVM and such.
In the end, ZFS is the single point that tempts me in general about Solaris, but I'm not about to jump platforms when I know enough 'tricks' to get 'good enough' out of my existing platform.
At least in linux, and I presume FreeBSD's swap strategy is similar, you miss the point. Let's look at two scenarios, one with proactive swapping, one without, and a malloc comes in that exceeds system memory.
Non-proactive case:
-kernel sees malloc, knows it lacks physical memory to accommodate, malloc is blocked while kernel does housekeeping.
-kernel picks the appropriate amount of pages to write to swap, then writes those pages to swap space, taking a while since block storage IO is excruciatingly slow.
-After the extremely long previous step, the memory is freed
-the malloc is allowed to continue, after a number of milliseconds have passed to execute the drive write, aside from the drive write, everything was in the microsecond scale, so it was delayed by a factor of thousands.
Proactive case:
-The system has some idle time, with nothing immediately better to do, the kernel notes free swap space and flags some appropriate memory as what would be swapped out if and when the system was in need, and copies it to disk, but it *leaves it in memory*. The kernel remembers that while these pages are indeed in memory, it can zap them and be able to restore. This is the critical point, the data in memory has not been *moved* to swap, it has been *copied* to swap.
-Program using that data randomly kicks back to life. It's needed data is on disk, but it is a moot point because it is in physical memory too, so it isn't slowed down. The kernel might take this opportunity to re-evaluate things when idle in terms of what it thinks is mostly unwanted pages.
-Later on, a program needs to malloc and physical memory is exhausted, the kernel blocks to do housekeeping, finds pages that it knows it has copied to disc, frees them and uses them to satisfy the malloc, within microseconds.
Proactive swapping causes extra IO activity during idle, but does not, if implemented correctly, impact things proactively swapped unnecessarily negatively, and allows swap on actual demand to be nearly trivially fast. It may be wasteful to have gobs of swap, and certainly if the swap has the sole copy of tons of data then performance is hopeless, but don't think seeing the swap used count go up 'mysteriously' without significant mallocs going on that it will impact access to the data written to swap later on.
The point is, the code from inception was organized by the developer as 'either GPL or BSD'. Anyone contributing code to that should recognize that. If someone choses to follow GPL, devs shouldn't get offended, they knew what they were getting into. If you don't want to play by the rules of a segment of a project, go away. Claiming that people cannot strip the BSD license on a redistribution is like saying because one of the licenses is GPL, you must always ship the source code. It's ok to dual-license something, but you have to recognize that each license explicitly grants rights that the other may preclude (i.e. BSD explicitly grants the right to rip code and use as you see fit, without source, which the GPL explicitly forbids). If a project choses to use the code under one license or the other, you simply shouldn't take offense. If you didn't want the code to be distributed under GPL, you shouldn't say it's ok in the first place. If it is more of a 'you must follow the rules of both', it kinda blows up and can't be used in either world. BSD as a whole is setup such that a third party should be able to use it with impunity, and code which may have the GPL apply precludes that. The same people bitching about BSD license being stripped would be happy to see the code redistributed as a binary.
Modern 'supercomputers' are not fundamentally so different from any other network of more than one server. Most run at least one independent kernel instance on each server, and that kernel is generally the same kernel as anyone else would use of that platform. The brains about sending messages efficiently between nodes is generally handled by userspace software (though you might make the claim that low-latency interconnects such as Infiniband and Myrinet are mostly cluster oriented, and have drivers that provide a low level medium for the userspace tools through their kernel drivers, but the story configuration involved none of that). To make something that no one could argue against having the same basic architecture as a 'real' cluster, you just need two-three computers with ethernet connectivity, and download something like OpenMPI and off you go.
Their number sounds about right for 4 2.0 ghz Opterons. If they used core2, they are doing pretty terrible at Rmax/Rpeak ratio. Maybe they have terribly low amounts of ram per core (which precludes tho Top500 benchmark from doing well).