Can anobody explain what are the benefits and drawbacks of an exokernel based system compared to a microkernel one? From what I understand, both are similar in that they have a minimal kernel. The difference is that microkernels use daemons to provide services to client programs while exokernels have the same functionality in user space libraries. At least exokernel advocates seem to be saying that exokernels are quicker than microkernels (and potentially quicker than traditional monolithic kernels too) because there is no need to context switch when you need system services. And microkernels need some kind of IPC to communicate with the daemons, which adds even more overhead, right?
Or the 10 MW used by the new ASCI Q computer Los Alamos is building. Well, it's a cluster, but still impressive.
Re:The thing I don't get about VLIW is this...
on
Itanium Problems
·
· Score: 2
IANEE either, but if you're running a sequential algorithm, OoO execution doesn't help that much. Neither does adding more cpu:s, for that matter. Of course, for OoO, there is a sweet spot beyond which it's not really worth adding execution units, and I suspect this limit is quite low. But what about SMT (symmetric multithreading)? Would it be possible to run different threads at the same time through different execution units?
Re:itanium is a solid chip from what I've seen...
on
Itanium Problems
·
· Score: 4, Interesting
Well, the steam generators don't have to be that big, actually. For example, there are steam locomotives in use which are about as powerful as similarly sized diesel locomotives, only their fuel consumption is a lot worse.
Anyway, you don't necessarily need steam either. There are those nuclear batteries used on spacecraft and shit like that. Terribly inefficient, but you get electricity from a nuclear reaction with no moving parts at all. And don't forget gas turbines, that many of the more modern nuclear powerstation designs are using. They can be a lot smaller than comparable steam generator systems. For example the Pebble bed modular reactor.
Re:itanium is a solid chip from what I've seen...
on
Itanium Problems
·
· Score: 2
Well, I think a place as large as google has their own datacenter. So they can custom design it any way they want. So, in the end, heat output is taken care of by more air conditioning, which raises the electrical bill. As an example, one of the US bomb labs is building a new computer where the computer uses 6 MW electricity and the cooling system another 4 MW.
Well, the chip still has to communicate with the outside world somehow. Optical interconnects are still quite far off. Come to think about it, so are 100 GHz chips.
MDK 9 uses KDE 3, which is supposed to be slightly faster than KDE 2.2.2 which is used in MDK 8.2. Also it uses gcc 3.2, which also is supposed to produce faster code than older gcc versions. So probably it will be faster. Noticably faster? Go figure.. If you want a noticable improvement in speed, change to a lightweight wm instead of heavy desktop environments like kde or gnome.
Problem is, for stuff like group calendaring and a lot of other stuff kroupware is going to do, I don't think there exists any standards. On the positive side, they are basing their stuff as much as possible on open standards, IIRC most of the non-standard stuff is rather straigh-forward MIME compliant extensions to IMAP. One can only hope they succeed, thus creating a standard for others to use.
But then again, in the Real World (TM), different processes/threads often need to communicate with each other (for ex. scientific applications), or save memory by sharing stuff like script interpreters, db connections etc. (for eg. web servers).
Umm, what's the point? The idea of a submarine is that it should be hard to detect. With that antenna sticking up from the water (or how are you going to control it) it kinda defeats the purpose.. although torpedoes would be übercool.
I think there are some 3rd party products implementing various kinds of clustering/HA/failover. For 7.3 (or was it 7.4) they are working to integrate replication into the core.
Scalability:
Well, postgres uses a multi-process model, like say, apache. So in principle it can scale quite well on an SMP system. Regarding clustering, I don't know if the current work on replication includes this or not. I'd guess that when you get replication working correctly, adding clustering is not a big deal. However, the kind of clustering were you have many servers working on the same data, like the oracle9 clustering, is still quite far off, I'd say
Secure data: postgres can do hot backups, yes. The pg_dump program outputs to standard output, so you can easily integrate it into any normal unix backup scheme with tape robots and whatnot.
Performance: postgres uses the OS file system, raw devices are not supported. So anything that the OS file system layer supports (e.g. raid) postgres supports. There was some talk about supporting raw devices, but it was decided that it was not worth the effort.
Well, if jetliners were fueled by hydrogen, the WTC towers would probably be standing today. Remember, the theory is that the burning kerosene weakened the steel support bars. If the planes had been fueled by hydrogen (which is a gas, and a very light one too) the hydrogen which wouldn't have been consumed in the initial explosion would have exited real quickly through the newly created ventilation holes in the building. Of course, this depends on how the hydrogen would be stored in the plane.
Actually, if there's room in the weight budget, you could keep all or part of the batteries as emergency reserve. It would be pretty compelling to have 100NM of reserve in the event of a fuel cell failure, though the motor itself seems far more likely to fail.
Umm, why? I'd say the fuel cell and it's support systems are much more likely to fail than the engine. While the article doesn't say, I don't think they used a DC engine with failure-prone brushes. A AC induction motor or a switched reluctance (SR) motor with assorted power semiconductors to control them is much more simpler, lightweight and efficient. Of course they could have used a permanent magnet motor too, but those are expensive as hell. Fuel cells, on the other hand, are a relatively unproven technology. There's lots of stuff there that could break. Maybe not the cell itself, but stuff like air compressors, fuel pumps etc.
Re:Compare and contrast
on
Gaim For Windows
·
· Score: 1, Offtopic
"I'll bet she gives great helmet." - Dark Helmet
Waaah! I just watched spaceballs (again)! Awesome flick...:)
As everyone else here, I don't know either. But I'd say it's quite a different kernel than the stock 2.4/2.5 kernel. I'd gues something like
1) A K42 -like exokernel with some parts of the linux kernel bolted on.
2) Something like Larry McVoys idea of OsLets, i.e. many kernels running on the system collaborating to provide a single system image to the user.
3) The traditional way, i.e. implementing super-fine-grained locking in the linux kernel. This would of course make linux hard to maintain and slow on "normal" hardware, just like say, solaris.
I've been doing ab initio calculations, i.e. calculating properties of some atomic system starting from quantum mechanics. Last month I used about 3000 CPU-hours of IBM POWER4 1.1Ghz juice. And my calculations weren't extremely complicated either..
It's the changing electric field that carries the information. The field propagates at about 2/3 times the speed of light in copper. The motion of the electrons is because of the field, not the other way around as you describe.
Hrm, you know, the computing world is bigger than office+quake+rendering. There are plenty of cpu hogging stuff besides rendering. But of course, your advice is entirely correct. Benchmark the stuff that's important for you. Not what all those hardware sites think that is important for their on average 14-year old readers.
The next version of enlightenment is supposed to use OpenGL for rendering. So it's a window manager thing, not a Xfree thing. Of course, it would be nice if Xfree would support hw rendering of all Xlib primitives.. Oh well..
Slashdot Mirror
Can anobody explain what are the benefits and drawbacks of an exokernel based system compared to a microkernel one? From what I understand, both are similar in that they have a minimal kernel. The difference is that microkernels use daemons to provide services to client programs while exokernels have the same functionality in user space libraries. At least exokernel advocates seem to be saying that exokernels are quicker than microkernels (and potentially quicker than traditional monolithic kernels too) because there is no need to context switch when you need system services. And microkernels need some kind of IPC to communicate with the daemons, which adds even more overhead, right?
Or the 10 MW used by the new ASCI Q computer Los Alamos is building. Well, it's a cluster, but still impressive.
IANEE either, but if you're running a sequential algorithm, OoO execution doesn't help that much. Neither does adding more cpu:s, for that matter. Of course, for OoO, there is a sweet spot beyond which it's not really worth adding execution units, and I suspect this limit is quite low. But what about SMT (symmetric multithreading)? Would it be possible to run different threads at the same time through different execution units?
Well, the steam generators don't have to be that big, actually. For example, there are steam locomotives in use which are about as powerful as similarly sized diesel locomotives, only their fuel consumption is a lot worse.
Anyway, you don't necessarily need steam either. There are those nuclear batteries used on spacecraft and shit like that. Terribly inefficient, but you get electricity from a nuclear reaction with no moving parts at all. And don't forget gas turbines, that many of the more modern nuclear powerstation designs are using. They can be a lot smaller than comparable steam generator systems. For example the Pebble bed modular reactor.
Well, I think a place as large as google has their own datacenter. So they can custom design it any way they want. So, in the end, heat output is taken care of by more air conditioning, which raises the electrical bill. As an example, one of the US bomb labs is building a new computer where the computer uses 6 MW electricity and the cooling system another 4 MW.
Or beaten by using a fork to remove the eyeball of some unsuspecting victim.
Well, the chip still has to communicate with the outside world somehow. Optical interconnects are still quite far off. Come to think about it, so are 100 GHz chips.
MDK 9 uses KDE 3, which is supposed to be slightly faster than KDE 2.2.2 which is used in MDK 8.2. Also it uses gcc 3.2, which also is supposed to produce faster code than older gcc versions. So probably it will be faster. Noticably faster? Go figure.. If you want a noticable improvement in speed, change to a lightweight wm instead of heavy desktop environments like kde or gnome.
Problem is, for stuff like group calendaring and a lot of other stuff kroupware is going to do, I don't think there exists any standards. On the positive side, they are basing their stuff as much as possible on open standards, IIRC most of the non-standard stuff is rather straigh-forward MIME compliant extensions to IMAP. One can only hope they succeed, thus creating a standard for others to use.
But then again, in the Real World (TM), different processes/threads often need to communicate with each other (for ex. scientific applications), or save memory by sharing stuff like script interpreters, db connections etc. (for eg. web servers).
Problem is, you'd need lots of servos to remote control a square rigged ship. Take a look at this page to get a picture of how complicated it is.
Umm, what's the point? The idea of a submarine is that it should be hard to detect. With that antenna sticking up from the water (or how are you going to control it) it kinda defeats the purpose.. although torpedoes would be übercool.
Availability:
I think there are some 3rd party products implementing various kinds of clustering/HA/failover. For 7.3 (or was it 7.4) they are working to integrate replication into the core.
Scalability:
Well, postgres uses a multi-process model, like say, apache. So in principle it can scale quite well on an SMP system. Regarding clustering, I don't know if the current work on replication includes this or not. I'd guess that when you get replication working correctly, adding clustering is not a big deal. However, the kind of clustering were you have many servers working on the same data, like the oracle9 clustering, is still quite far off, I'd say
Secure data:
postgres can do hot backups, yes. The pg_dump program outputs to standard output, so you can easily integrate it into any normal unix backup scheme with tape robots and whatnot.
Performance:
postgres uses the OS file system, raw devices are not supported. So anything that the OS file system layer supports (e.g. raid) postgres supports. There was some talk about supporting raw devices, but it was decided that it was not worth the effort.
Actually, since version 7.1 (current is 7.2) row size is unlimited. Or, rather the limits are imposed by the operating system (2GB files on ext2?).
Well, if jetliners were fueled by hydrogen, the WTC towers would probably be standing today. Remember, the theory is that the burning kerosene weakened the steel support bars. If the planes had been fueled by hydrogen (which is a gas, and a very light one too) the hydrogen which wouldn't have been consumed in the initial explosion would have exited real quickly through the newly created ventilation holes in the building. Of course, this depends on how the hydrogen would be stored in the plane.
Actually, if there's room in the weight budget, you could keep all or part of the batteries as emergency reserve. It would be pretty compelling to have 100NM of reserve in the event of a fuel cell failure, though the motor itself seems far more likely to fail.
Umm, why? I'd say the fuel cell and it's support systems are much more likely to fail than the engine. While the article doesn't say, I don't think they used a DC engine with failure-prone brushes. A AC induction motor or a switched reluctance (SR) motor with assorted power semiconductors to control them is much more simpler, lightweight and efficient. Of course they could have used a permanent magnet motor too, but those are expensive as hell. Fuel cells, on the other hand, are a relatively unproven technology. There's lots of stuff there that could break. Maybe not the cell itself, but stuff like air compressors, fuel pumps etc.
"I'll bet she gives great helmet." - Dark Helmet
Waaah! I just watched spaceballs (again)! Awesome flick...:)
As everyone else here, I don't know either. But I'd say it's quite a different kernel than the stock 2.4/2.5 kernel. I'd gues something like
1) A K42 -like exokernel with some parts of the linux kernel bolted on.
2) Something like Larry McVoys idea of OsLets, i.e. many kernels running on the system collaborating to provide a single system image to the user.
3) The traditional way, i.e. implementing super-fine-grained locking in the linux kernel. This would of course make linux hard to maintain and slow on "normal" hardware, just like say, solaris.
I've been doing ab initio calculations, i.e. calculating properties of some atomic system starting from quantum mechanics. Last month I used about 3000 CPU-hours of IBM POWER4 1.1Ghz juice. And my calculations weren't extremely complicated either..
EXPENSIVE
I have 100Mbps. At home. For about EUR 10 per month, IIRC. One of the benefits of uni housing...:)
It's the changing electric field that carries the information. The field propagates at about 2/3 times the speed of light in copper. The motion of the electrons is because of the field, not the other way around as you describe.
I don't know what to think about this. Of course it's nice that real scientists get their work done, but on the other hand:
"In space, noone can hear you sing."
Hrm, you know, the computing world is bigger than office+quake+rendering. There are plenty of cpu hogging stuff besides rendering. But of course, your advice is entirely correct. Benchmark the stuff that's important for you. Not what all those hardware sites think that is important for their on average 14-year old readers.
The next version of enlightenment is supposed to use OpenGL for rendering. So it's a window manager thing, not a Xfree thing. Of course, it would be nice if Xfree would support hw rendering of all Xlib primitives.. Oh well..