| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
Comments:
<0> oops.
<0> yeah.
<0> 25 hours for 1 drive for 8tb.
<1> l8r
<0> ***uming instantaneous tape changes.
<0> so 10 drives would be about 2.5 hours.
<0> and no need to change tapes.
<2> lowkey: true. and faster robot != more reliable system. hehehe.
<3> you guys are all dreaming
<0> still far beyond the capabilities of the average company or consumer.
<2> lowkey: slower robot = system lasts longer. tapes don't get fscked
<0> 40gb is hard enough for most to backup.
<2> lowkey: so d2d really is awesome. lol
<0> and thats only 8-10 dvds.
<4> how do you scp a file from 1 remote machine to another when both machines have ssh running on non-standard ports? ie: box1:file.txt port 1234 ---> box2:file.txt port 1234
<0> unless you're running ssh agent on all three boxes that wont work.
<5> litage: i'd ***ume scp -P port1 user@host1:files -P port2 user@host2:.
<2> lowkey: if you have fistfulls of money, the best way to run your shop is to buy multiple magnetic storage vendors products and mirror from one to the other. hehehe. that way, if one breaks, the other is still up. :->
<6> Ello
<3> most companies are small businesses....most companies print reports and send them to their accountant....most companies dont do any data analysis....most companies have almost zero it knowledge
<0> linagee: on any large install, you hit failure rates like clockwork.
<2> lowkey: naw. :-P
<2> lowkey: :)
<3> and most companies will always be that way ;)
<2> lowkey: hehehe
<0> linagee: whats the biggest thing you've admined?
<0> linagee: i used to work clusters.
<2> lowkey: start putting 100 drives in a system and one will fail about every week from what i've experienced. :o)
<2> lowkey: i've probably touched several petabytes of data. lol
<0> with 100 anyting something will fail every week.
<7> linagee: what was the purpose of the clusters?
<0> arrummzen: me?
<2> lowkey: the newest system i work on has 480 drives. :-)
<0> arrummzen: small ones for accountants.
<2> lowkey: muahhha! :o)
<0> arrummzen: large ones for weather analysis.
<7> hmmm...
<8> http://img158.imageshack.us/img158/626/screen4ve.jpg
<2> lowkey: i bet you have to feed it a drive every week at least! heh
<0> linagee: damn never worked storage that large.
<0> linagee: i worked computational clusters.
<2> lowkey: imagine you use 500GB drives (which is supported by the system)
<8> input?
<0> mostly network card and mainboards failed.
<2> lowkey: 480 * 500
<0> linagee: most nodes need less than 1gb so we used the smallest drives we could find cheap.
<2> lowkey: scrap that. the first tray must be FC. (480 - 15) * 500
<2> lowkey: oh and actual usable space is probably like 480 gigs.
<0> linagee: the master nodes would access a RAID for the storage.
<0> and its size was dependent on the job.
<2> lowkey: any storage array?
<0> once.
<0> not commonly.
<9> hello
<2> lowkey: or only qualified vendors? heh
<9> does anyone knows how to fix the grub error 15 ?
<10> Jatt_kol_12_borh:
<0> these guys wanted processor power. so all money went to that.
<10> Jatt_kol_12_borh: Not me. Read the manual thouroughly?
<0> biggest i worked was a 768 node beowulf.
<9> i did i am in a bad situation right now
<2> lowkey: interesting. is it an AMD cluster? if so, i might know the company that owns it... :o()
<2> :o)
<11> *claps*
<9> i have hda <-- Windows and hdb <- Linux
<0> no that one was a pentium xeon duals for each node.
<9> now i cant boot neither
<0> quads for the masters.
<11> I got a new comp ^_^
<2> lowkey: FC backend?
<11> got a linmodem working too.. *wipes forehead*
<2> lowkey: er, FC interconnect between nodes?
<0> linagee: myrinet.
<0> linagee: over fiber.
<2> lowkey: nevermind. never touched the stuff
<11> mm fiber
<2> lowkey: it's weird how few people own myrinet. (or i've just never seen it)
<0> linagee: faster than ****. insanely low latency. pci cards fail regularly.
<0> daily.
<2> bsdirl: fiber is good for your diet. :)
<5> Jatt_kol_12_borh: http://www.gentoo.org/doc/en/grub-error-guide.xml
<0> hourly some days.
<2> lowkey: wtf?!?!
<0> they had issues at the time. mostly resolved now
<2> lowkey: the card costs like a thousand. why would you buy a super fast card if it costs so much and fails so often? lol
<0> cause it was warrantied.
<11> I ate my greens today, so I'm good :D
<0> and it was that low a latency.
<2> lowkey: aha. i see...
<0> although not as low as dolphin ics.
<0> those are cool.
<0> direct memory injection on remote nodes.
<2> ????
<12> Anyone have any experience with NPI's Keystone24G switch?
<0> direct access to RAM on remote nodes.
<2> lowkey: you say injection... what do you mean....
<0> no service.
<0> lower latency.
<2> lowkey: and why can't you do this with myrinet or something and the appropriate drivers
<0> they try to.
<0> it just doesn't work as well.
<0> plus dolphin uses a 3d model
<2> lowkey: you mean a matrix?
<0> so any two nodes are no more than two hops from each other.
<2> lowkey: what is the physical interconnect? copper? fiber?
<13> I've read recommendations in lots of places to use Qt, but it seems to be discouraged by some people
<13> do you guys know why?
<0> and you have at least 6 different paths to any other host so you can route around congestion.
<13> only ***ociation I can make with Qt that ****s is probably KDE
<0> linagee: i've only used dolpin ics on copper but they have fiber as i recall.
<2> lowkey: "congestion" lol. "that path is 5 nanoseconds too slow! blech!"
<0> exactly.
<5> excelblue: it works for me
<13> alright, so there's no reason not to use it?
<2> lowkey: i have yet to see an interconnect that works at memory speed. :-P
<2> lowkey: er, FSB speed.
<0> well fastest it can go is the bus speed if it was 100% efficient.
<0> ***uming a perfect world.
<5> excelblue: there's a religious war between the qt and gtk folks.. qt is c++ and gtk is plain c.. qt has much nicer file dialogs though, and the signals/slots model seems nice too
<2> lowkey: use a card and the fastest it can go is that bus speed. lol. (normal PCI, 66Mhz, etc.)
<2> lowkey: sorry, 33Mhz. or whatever
<0> well you use these in pci-x slots.
<0> generally.
<2> lowkey: PCI-X ****s. :-P
<0> but its fast.
<2> lowkey: it's only available on high end servers
<0> yep.
<2> lowkey: start using PCI-express already. :-P
<0> and its fast.
<0> pci-x is faster.
<11> aww the 3d accel. for my card ****s in this distro :[
<2> lowkey: PCI express is faster. heh
<0> sure about that?
<2> lowkey: faster, 133Mhz PCI-X 64bit or PCI express 16 or 8 channel connection
<2> lowkey: keep in mind not many motherboards and/or cards can use 133Mhz and just jump back down to 66Mhz. :-P
<0> ok.
<0> most pci-x is slower than pci-express
<0> but pci-x 266 is as fast
<0> and pci-x 533 is faster.
<2> 266? wtf. why.
<2> 533???
<0> http://www.it-enquirer.com/main/ite/printpage/161/
<2> why
<2> kill it. lol
<0> even faster than 4x infiniband.
<2> lowkey: 16 channel pci express is hellafast
<0> yeah.
<0> i bet it is.
<2> lowkey: most graphics cards don't even touch it's capabilities
<2> er, s/most/all/
Return to
#linux
or
Go to some related
logs:
#linuxhelp
file size limit exceeded unrar
fc5 increase file descriptors
ubuntu autohinter etch
#perl
#perl
xubuntu not graphical
#gimp
#perl
#python
|
|
