This guy's right... Netapps are RAID-4. (Or at least, the older ones are. They might be using some kind of double-parity RAID-4ish thing now...) According to the CMU Raidframe papers, RAID-6 is similar to RAID-5, but with two (or more) paritys split across multiple drives.
The dedicated parity disk(s) aren't a bottleneck with the Netapp, since you've got all that nvram -- things won't actually get written out to disk until there's a full stripe to write.
Everybody here seems to be jumping on the hardware RAID bandwagon. Why? The "software RAID has too much overhead" arguement is bogus -- parity calculations are cheap compared to the amount of time it takes to actually write stuff to disk. Modern busses like SCSI aren't terribly CPU intensive, either. And more often than not, it's I/O that's the bottleneck -- not CPU.
I also have to reiterate what other people have been saying: just because the drives are flashing all the time doesn't mean those disks are saturated.
Does Solaris do softupdates? Does sun offer a prestoserve (NVRAM) card? Especially with lots of files which are changed rapidly (which happens on a mail server) both softupdates and prestoserve could save a lot of that I/O from even hitting the disk in the first place...
If you do want to go RAID, I'd suggest RAID0+1. (mirror the individual stripes, instead of striping the mirrors.) It eats disks, but gives you the potential of losing half your drives without hurting anything. Split the mirrors across two cabinets with two separate SCSI busses, and if one whole cabinet dies, you still come out smelling like roses.
If you've dealt with NetBSD at all, you'd know that things don't get done unless they're done in an architecture-independent manner. Until XFree86 86s the x86 and uses sane bus abstractions instead of counting on directly diddling the PCI bus, nobody [1] from the NetBSD team is terribly interested in supporting it. The XFree86 team knows that architecture-independence is the way to go, and supposedly v4.0 will be implemented a lot more cleanly.
[1] There are actually some NetBSD guys who'd rather have ``impure'' XFree86 support than none at all, and are working on getting it running under NetBSD/alpha.
NetBSD is production-class on many alphas, but not all of them. OSF^WDU^WTru64 also supports more of the whacky proprietary DEC hardware (like turbochannel graphics cards) than NetBSD does at the moment... I'm sure Linux is in the same boat with its Alpha support.
Having a working box with a proprietary OS today is sometimes preferable to having a non-working box with a soon-to-come free OS.
Also does this unix have an install tool like rpm?
yes, although I forget the name at the moment.
Also, well for the most of us alpha processor systems are toooo expensive? Anyone know where to get a sub $1500 system, that can handle (if not have) multiple processors.
I got my two 3000/400s free, since a local university was going to throw them out, and I know one of the head sysadmins there. Remember that some of the early alphas are now five or six years old, and are considered ``junk'' by a lot of universities and businesses. Sure, a PIII may toast my piddly 133Mhz 20164, but floating point on even the old alphas is still pretty impressive. Plus there's the whole 64-bit thing.
As for multiprocessor machines, you just might get lucky!
It's like women are not even people to the author... just some kind of ``pleasure automata'' to serve, not dissimilar to those computers which give so much satisfaction. A few choice quotes:
We need what are now called "old fashioned girls" who don't mind cooking our meals, rubbing our sore shoulders, and running our bath water for us.
``I want a wife who'll give me a blowjob, rub my feet, and have dinner ready for me when I get home from a long day of hacking. I'd like a wife who is my personal slave, because, hey, that's all marriage really is.''
Re Hot Babes:[...]
I generally found them to be more trouble than they were worth. [...]The trick is to find a woman who doesn't spend a lot of time and money cuting herself up, but is pleasant to hold once all the packaging is removed. She'll be more likely to want some cuddling than the vain ones, and, unlike them, will concentrate on loving you instead of worrying about getting her hair messed up. [...]The best software usually doesn't come in the fanciest box, right? The same goes for girls.
How's that for a big sweeping generalization, (All glamorous women ``aren't worth it,'') coupled with objectification. (Women == software.)
And if the girl finds you unattractive, she'll let you know that, too (so you can dump her before you get too serious).
Yeah, because really the issue here is how male geeks can get chicks without changing any aspect of themselves! Obviously our author can't think outside of his box...
Understanding women is harder than figuring out the hardest computer game, harder even than setting up a secure 200-client network running *BSD. But women can offer more satisfaction than even an overclocked, dual-Celeron workstation, so learning how to deal with them is worth the extra effort.
``But you know, no woman would stand a chance of giving the same satisfaction as running a beowulf cluster,'' as if women and computers are even remotely comparable in some way.
[M]
y wife likes to redecorate frequently, which sometimes annoys me, but I've learned to shrug my shoulders and call this part of her personality a feature, not a bug[.]
``In fact, I originally though about GPLing our marriage, but wasn't too fond of the free distribution clause, so settled on a proprietary fork of BSD.''
The primary trouble with geeks meeting people is not so much any misunderstandings of social etiquette and (possibly lack of) associated skills, but the unwillingness to stop thinking of the external world in terms of a computer-related mindset. The above article does nothing to persuade its readers to move beyond the ``all the world's a computer and we are but its processes'' mentality, yet ditching such a fixed and limited schema is imperative not only for meeting other people who don't share such a schema, but for personal growth and maturation.
People (including women) aren't automata, there's no ``system'' for ``getting'' them, and instead of analyzing people and trying to find the perfect one, analyze yourself and change into the person who will attract people you're interested in.
The ev6 beat my humble 366c but I'm afraid my friend's 500 MHz Celeron beats it soundly. I don't remember the numbers he told me yesterday, but I remember it beating my 366 by 25% or so.
This proves nothing until you profile the code. It could be that your random number generator eats more CPU time on alpha for whatever reason... You also neglected to mention what compiler you were using -- gcc on alpha is not terribly efficient.
Funny how many of the intermediate processors aren't being fabbed anymore. Anyone still making 8086s (acutally, probably) 286s? 386s? 486s?
Usually what happens (with intel at least) is that they modify their old designs and make ``embedded'' versions of the chips in their older fabs. The company I work for (who shall remain nameless) uses the 80186 (no joke) and the i386EX (32-bit mode only 80386) in two of its products.
There, I've just forwarded in my mail server. It works with udp as well. You can also add multiple machines and round-robin between them. Do port translation to get around firewalls. Forward out different services to different machines and confuse the hell out of skript kiddies.
Protocols which break behind NAT boxes are doing nasty stuff like including IP and port numbers inside the _data_ layer of the packet, which is arguably a stupid thing to do.
That new PlayStation will be what the real Amiga would have been if it had survived.
Never mind that sony won't give you docs on the thing without signing an NDA, whereas hardware information on the Amiga was free and plentiful, hence all the nifty demos. Even if sony does the Net Yaroze thing with the new playstation, you can bet your pants that it'll be the same old black box (no pun intended) routines without true hardware docs.
The playstation is just a super-nifty-whiz-bang console game playing machine. The Amiga was an artist's tool.
I would like to see linux go the microkernel direction (or atleast have the choice-micro or mono when you compile the thing).
Linus himself says that microkernels are a sham. But if you really want to run linux on top of a microkernel, that's what mkLinux is all about.
Or even better, make linux an exokernel, that beats them all out.
Yeah, never mind the fact that as soon as you have to decide policy, you have to start implementing what basically amounts to a microkernel. The exokernel idea is rather neat from a theoretical standpoint, but simply not workable in practice, at least not without some hardware help. (OS/390 anybody?)
Everybody wants to write k-rAd widget sets, toolboxes, and design themes for the most bloated window manager that ever existed, but who's working on the apps?
What no new ideas? Damn it! Must we wait for Apple for *ALL* GUI innovation?
Of course not. NeXTstep still kicks ass in terms of usability and interface, while still being unix underneath. I don't know why the hell everybody's jumping on the gnome and KDE bandwagons when GNUStep is around.
HTTP 1.0 required a new stream for each packet of data sent. But HTTP 1.1 can send multiple packets along the same stream, speeding the flow of information on the Web.
A new TCP stream for each packet? No wonder the world wide web is so slow!:P
I wonder if it's hopeless to think the media in general can get terminologies correct.
IDE has no notion of "disconnect" like SCSI, so the bus is held for the full duration of a read or write, which limits the usable full bus bandwidth to one device per bus. Technically there's no reason why you couldn't run a RAID5 on three or four IDE devices on two IDE busses, but it isn't practical since you're basically halving your bandwidth per device.
This shouldn't stop you from playing around, though... I once made a RAID0 with two old 80mb IDE drives on the same bus. It was slower than a single drive, but had twice the capacity.:)
2) Nearby street lights broadcast the signal Um... this is just plain silly. But the signal can be snooped by your neighbors who share your transformer. So we encrypt the data. Problem solved.
You miss the obvious -- that all this excess radiation is going to cause interference all over the place. It's bad enough that a 50/60Hz hum and associated harmonics are always showing up in wires everywhere due to the lack of shielding in powerlines and house wiring, but now you want to spew even more RMI throughout the spectrum with your networking-over-powerlines hack?
Powerlines were never designed to carry bandwidths required for transmitting data signals. As such, they lack fundamental safeguards against RMI such as shielding and are thus unsuitable for data networking.
For content, sure... but this fscking web interface sucks. Good ol' trn, last updated in 1993, still kicks the pants off of the scroll and drool web interface for usability in reading large numbers of messages.
there is only one mainstream 'thread' of the linux sources
Ahh, but there is more to a complete linux system than the kernel alone. The kernel itself is absolutely useless without other supporting programs like a shell at the very least. Witness the ``many different versions'' of distributions the linux community has come up with. This is confusing for end-users, and for developers alike.
MP3 isn't open. As I understood it, legally there's nothing stopping Fraunhofer from ordering an injunction or something similar and declaring that all MP3 encoders and decoders are covered by their patent(s). (In fact, I believe they've done so, but aren't enforcing it...)
Any ``standards'' propigated by businesses with pecuniary interest in that standard aren't ``open'' by any means. The original architects of the internet were working off of government grants, not off of the idea that what they were creating would make them rich later. It'd be nice if a large body like the FSF could similarly fund research into truly open standard formats.
Besides, there's better ways to analyze music than DFT's, wavelets being one method...
It's bad enough that they integrated IDE controller with motherboards
Blame intel for the IDE integration... they make most of the IDE controller chips, and almost all of the first-generation motherboards are designed by intel. IDE has come a long way in the last couple years, but SCSI it's still not.
what are you gonna do later when you want to upgrade your machine?
Disable the onboard stuff, slap your new whizbang video card in, and be happy. Or else just slap your new whizbang video card in, and have _two_ heads on your machine, or _two_ soundcards, or _two_ whatevers...
I am still using my SB16 that I originally bought with my 486 over 3 years ago.
And it's still a sucky SB16. Go figure. If you had gotten an integrated motherboard, you'd have gotten better sound and video.
Same goes for those "integrated" SCSI and network cards.
I suppose you also think that all those manufacturers of real unix workstations are stupid dolts for integrating stuff like this on the motherboard. Duh. Thank god people like you never design real workstations.
Slashdot Mirror
``I run NetBSD as my OS of choice, and I still read /.''
This guy's right... Netapps are RAID-4. (Or at least, the older ones are. They might be using some kind of double-parity RAID-4ish thing now...) According to the CMU Raidframe papers, RAID-6 is similar to RAID-5, but with two (or more) paritys split across multiple drives.
The dedicated parity disk(s) aren't a bottleneck with the Netapp, since you've got all that nvram -- things won't actually get written out to disk until there's a full stripe to write.
Everybody here seems to be jumping on the hardware RAID bandwagon. Why? The "software RAID has too much overhead" arguement is bogus -- parity calculations are cheap compared to the amount of time it takes to actually write stuff to disk. Modern busses like SCSI aren't terribly CPU intensive, either. And more often than not, it's I/O that's the bottleneck -- not CPU.
I also have to reiterate what other people have been saying: just because the drives are flashing all the time doesn't mean those disks are saturated.
Does Solaris do softupdates? Does sun offer a prestoserve (NVRAM) card? Especially with lots of files which are changed rapidly (which happens on a mail server) both softupdates and prestoserve could save a lot of that I/O from even hitting the disk in the first place...
If you do want to go RAID, I'd suggest RAID0+1. (mirror the individual stripes, instead of striping the mirrors.) It eats disks, but gives you the potential of losing half your drives without hurting anything. Split the mirrors across two cabinets with two separate SCSI busses, and if one whole cabinet dies, you still come out smelling like roses.
If you've dealt with NetBSD at all, you'd know that things don't get done unless they're done in an architecture-independent manner. Until XFree86 86s the x86 and uses sane bus abstractions instead of counting on directly diddling the PCI bus, nobody [1] from the NetBSD team is terribly interested in supporting it. The XFree86 team knows that architecture-independence is the way to go, and supposedly v4.0 will be implemented a lot more cleanly.
[1] There are actually some NetBSD guys who'd rather have ``impure'' XFree86 support than none at all, and are working on getting it running under NetBSD/alpha.
NetBSD is production-class on many alphas, but not all of them. OSF^WDU^WTru64 also supports more of the whacky proprietary DEC hardware (like turbochannel graphics cards) than NetBSD does at the moment... I'm sure Linux is in the same boat with its Alpha support.
Having a working box with a proprietary OS today is sometimes preferable to having a non-working box with a soon-to-come free OS.
yes, although I forget the name at the moment.
I got my two 3000/400s free, since a local university was going to throw them out, and I know one of the head sysadmins there. Remember that some of the early alphas are now five or six years old, and are considered ``junk'' by a lot of universities and businesses. Sure, a PIII may toast my piddly 133Mhz 20164, but floating point on even the old alphas is still pretty impressive. Plus there's the whole 64-bit thing.
As for multiprocessor machines, you just might get lucky!
It's like women are not even people to the author... just some kind of ``pleasure automata'' to serve, not dissimilar to those computers which give so much satisfaction. A few choice quotes:
``I want a wife who'll give me a blowjob, rub my feet, and have dinner ready for me when I get home from a long day of hacking. I'd like a wife who is my personal slave, because, hey, that's all marriage really is.''
How's that for a big sweeping generalization, (All glamorous women ``aren't worth it,'') coupled with objectification. (Women == software.)
Yeah, because really the issue here is how male geeks can get chicks without changing any aspect of themselves! Obviously our author can't think outside of his box...
``But you know, no woman would stand a chance of giving the same satisfaction as running a beowulf cluster,'' as if women and computers are even remotely comparable in some way.
``In fact, I originally though about GPLing our marriage, but wasn't too fond of the free distribution clause, so settled on a proprietary fork of BSD.''
The primary trouble with geeks meeting people is not so much any misunderstandings of social etiquette and (possibly lack of) associated skills, but the unwillingness to stop thinking of the external world in terms of a computer-related mindset. The above article does nothing to persuade its readers to move beyond the ``all the world's a computer and we are but its processes'' mentality, yet ditching such a fixed and limited schema is imperative not only for meeting other people who don't share such a schema, but for personal growth and maturation.
People (including women) aren't automata, there's no ``system'' for ``getting'' them, and instead of analyzing people and trying to find the perfect one, analyze yourself and change into the person who will attract people you're interested in.
There's much more to life than computers and /.
This proves nothing until you profile the code. It could be that your random number generator eats more CPU time on alpha for whatever reason... You also neglected to mention what compiler you were using -- gcc on alpha is not terribly efficient.
Usually what happens (with intel at least) is that they modify their old designs and make ``embedded'' versions of the chips in their older fabs. The company I work for (who shall remain nameless) uses the 80186 (no joke) and the i386EX (32-bit mode only 80386) in two of its products.
I'll take nixie tubes over LEDs anyday...
Nothing satisfies quite like the scent of baked dust and the warm glow of vacuum tubes.
ipmasqadm portfw -a -P tcp -L 25 -R 25
There, I've just forwarded in my mail server. It works with udp as well. You can also add multiple machines and round-robin between them. Do port translation to get around firewalls. Forward out different services to different machines and confuse the hell out of skript kiddies.
Protocols which break behind NAT boxes are doing nasty stuff like including IP and port numbers inside the _data_ layer of the packet, which is arguably a stupid thing to do.
Anthony is a real guy. I went to high school with him.
Never mind that sony won't give you docs on the thing without signing an NDA, whereas hardware information on the Amiga was free and plentiful, hence all the nifty demos. Even if sony does the Net Yaroze thing with the new playstation, you can bet your pants that it'll be the same old black box (no pun intended) routines without true hardware docs.
The playstation is just a super-nifty-whiz-bang console game playing machine. The Amiga was an artist's tool.
Linus himself says that microkernels are a sham. But if you really want to run linux on top of a microkernel, that's what mkLinux is all about.
Yeah, never mind the fact that as soon as you have to decide policy, you have to start implementing what basically amounts to a microkernel. The exokernel idea is rather neat from a theoretical standpoint, but simply not workable in practice, at least not without some hardware help. (OS/390 anybody?)
Everybody wants to write k-rAd widget sets, toolboxes, and design themes for the most bloated window manager that ever existed, but who's working on the apps?
Of course not. NeXTstep still kicks ass in terms of usability and interface, while still being unix underneath. I don't know why the hell everybody's jumping on the gnome and KDE bandwagons when GNUStep is around.
A new TCP stream for each packet? No wonder the world wide web is so slow! :P
I wonder if it's hopeless to think the media in general can get terminologies correct.
IDE has no notion of "disconnect" like SCSI, so the bus is held for the full duration of a read or write, which limits the usable full bus bandwidth to one device per bus. Technically there's no reason why you couldn't run a RAID5 on three or four IDE devices on two IDE busses, but it isn't practical since you're basically halving your bandwidth per device.
This shouldn't stop you from playing around, though... I once made a RAID0 with two old 80mb IDE drives on the same bus. It was slower than a single drive, but had twice the capacity. :)
You miss the obvious -- that all this excess radiation is going to cause interference all over the place. It's bad enough that a 50/60Hz hum and associated harmonics are always showing up in wires everywhere due to the lack of shielding in powerlines and house wiring, but now you want to spew even more RMI throughout the spectrum with your networking-over-powerlines hack?
Powerlines were never designed to carry bandwidths required for transmitting data signals. As such, they lack fundamental safeguards against RMI such as shielding and are thus unsuitable for data networking.
For content, sure... but this fscking web interface sucks. Good ol' trn, last updated in 1993, still kicks the pants off of the scroll and drool web interface for usability in reading large numbers of messages.
Pah! VMS had clustering back in the early 80s.
Ahh, but there is more to a complete linux system than the kernel alone. The kernel itself is absolutely useless without other supporting programs like a shell at the very least. Witness the ``many different versions'' of distributions the linux community has come up with. This is confusing for end-users, and for developers alike.
MP3 isn't open. As I understood it, legally there's nothing stopping Fraunhofer from ordering an injunction or something similar and declaring that all MP3 encoders and decoders are covered by their patent(s). (In fact, I believe they've done so, but aren't enforcing it...)
Any ``standards'' propigated by businesses with pecuniary interest in that standard aren't ``open'' by any means. The original architects of the internet were working off of government grants, not off of the idea that what they were creating would make them rich later. It'd be nice if a large body like the FSF could similarly fund research into truly open standard formats.
Besides, there's better ways to analyze music than DFT's, wavelets being one method...
Of course I know it...
:)
I wrote one of the songs in it.
Blame intel for the IDE integration... they make most of the IDE controller chips, and almost all of the first-generation motherboards are designed by intel. IDE has come a long way in the last couple years, but SCSI it's still not.
Disable the onboard stuff, slap your new whizbang video card in, and be happy. Or else just slap your new whizbang video card in, and have _two_ heads on your machine, or _two_ soundcards, or _two_ whatevers...
And it's still a sucky SB16. Go figure. If you had gotten an integrated motherboard, you'd have gotten better sound and video.
I suppose you also think that all those manufacturers of real unix workstations are stupid dolts for integrating stuff like this on the motherboard. Duh. Thank god people like you never design real workstations.