how to start with own Beowulf

Utpal Jyoti Barman Utpal Jyoti Barman
Sun Aug 26 22:26:03 PDT 2001


Hii there,
 i am very new to Distributed Computing. Could anyone please help me to proceed with the Beowulf Architecture. I want to make one in my Lab consisting of 6 nodes.

With warm regards
Utpal


On Sun, 26 Aug 2001 beowulf-admin at beowulf.org wrote :
>
>Send Beowulf mailing list submissions to
>	beowulf at beowulf.org
>
>To subscribe or unsubscribe via the web, visit
>	http://www.beowulf.org/mailman/listinfo/beowulf
>or, via email, send a message with subject or body 
>'help' to
>	beowulf-request at beowulf.org
>You can reach the person managing the list at
>	beowulf-admin at beowulf.org
>
>When replying, please edit your Subject line so it is 
>more specific than
>"Re: Contents of Beowulf digest..."
>
>
>Today's Topics:
>
>  1. "mmap failed"---help (pramod ps)
>  2. Re:"mmap failed"---help (Sean Dilda)
>  3. RE:Network RAM for Beowulf (Thomas R Boehme)
>  4. Re:HELP (Brian)
>  5. Re:HELP (Ron Chen)
>  6. Re:HELP (=?iso-8859-1?Q?Jakob_=D8stergaard?=)
>  7. Re:Network RAM for Beowulf 
>(=?iso-8859-1?Q?Jakob_=D8stergaard?=)
>  8. Re:Network RAM for Beowulf 
>(=?iso-8859-1?Q?Jakob_=D8stergaard?=)
>  9. Re:Network RAM for Beowulf (Mark Hahn)
>  10. Re:HELP (Mark Hahn)
>  11. Re:Need to do something Useful (Kim Branson)
>  12. Re:Network RAM for Beowulf (andreas boklund)
>  13. Re:Network RAM for Beowulf (Eray Ozkural (exa))
>
>--__--__--
>
>Message: 1
>Date: Sat, 25 Aug 2001 10:13:56 -0700 (PDT)
>From: pramod ps 
>Subject: "mmap failed"---help
>To: beowulf at beowulf.org
>
>hi there,
>
>  i tried to impliment the beowulf cluster using scyld
>cd in my college. i've instaled the server & booted
>the clients using floppy disks.
>then i partitioned the hard disk of the client
>remotely by giving commands, beofdisk -d & beofdisk
>-w.
>
>  then i restart the machine , edit the fstab &tried
>to execute a program. but the client showed an error
>message as , " mmap failed".
>when the program is given to execute in the server ,
>it succesfully executed & showed the result. i'am
>using only one client to try this.
>
>  its possible to create & write the partition table
>to client remotely. but when we tried to write the
>boot record & boot the system from hard disk , it
>showed that the there is no bootable partition. 
>i think the problem is that of mounting the client
>hard disk to the server.
>
>   i've to complete this with in the next week. so
>kindly help me as early as possible to clear this
>problem.
>
>thank u ,
>pramod.  
>
>__________________________________________________
>Do You Yahoo!?
>Make international calls for as low as $.04/minute with 
>Yahoo! Messenger
>http://phonecard.yahoo.com/
>
>--__--__--
>
>Message: 2
>Date: Sat, 25 Aug 2001 13:26:04 -0400
>From: Sean Dilda 
>To: pramod ps 
>Cc: beowulf at beowulf.org
>Subject: Re: "mmap failed"---help
>protocol="application/pgp-signature"; 
>boundary="qMm9M+Fa2AknHoGS"
>
>
>--qMm9M+Fa2AknHoGS
>Content-Type: text/plain; charset=us-ascii
>Content-Disposition: inline
>Content-Transfer-Encoding: quoted-printable
>
>On Sat, 25 Aug 2001, pramod ps wrote:
>
>> hi there,
>>=20
>>   i tried to impliment the beowulf cluster using scyld
>> cd in my college. i've instaled the server & booted
>> the clients using floppy disks.
>> then i partitioned the hard disk of the client
>> remotely by giving commands, beofdisk -d & beofdisk
>> -w.
>>=20
>>   then i restart the machine , edit the fstab &tried
>> to execute a program. but the client showed an error
>> message as , " mmap failed".
>
>Can you give me more details about where exactly you 
>saw the "mmap
>failed" error message
>
>> when the program is given to execute in the server ,
>> it succesfully executed & showed the result. i'am
>> using only one client to try this.
>>=20
>>   its possible to create & write the partition table
>> to client remotely. but when we tried to write the
>> boot record & boot the system from hard disk , it
>> showed that the there is no bootable partition.=20
>> i think the problem is that of mounting the client
>> hard disk to the server.
>
>The master node never mounts the slave nodes harddrives.
>
>NOTE: I assume by 'server' you mean master node, and by 
>'client' you
>mean slave node or compute node.. please let me know if 
>I am wrong in
>that.
>
>--qMm9M+Fa2AknHoGS
>Content-Type: application/pgp-signature
>Content-Disposition: inline
>
>-----BEGIN PGP SIGNATURE-----
>Version: GnuPG v1.0.1 (GNU/Linux)
>Comment: For info see http://www.gnupg.org
>
>iD8DBQE7h9+s+pq97aGGtXARAtr2AKCFLjxqdpE4ZgHPwc7Dvl0FQp7Z-
>zgCeL75e
>KdiMu3+xFg5pqgkY3J9Xvss=
>=dd2V
>-----END PGP SIGNATURE-----
>
>--qMm9M+Fa2AknHoGS--
>
>--__--__--
>
>Message: 3
>From: Thomas R Boehme 
>To: "'Mark Hahn'" , 
>beowulf at beowulf.org
>Subject: RE: Network RAM for Beowulf
>Date: Sat, 25 Aug 2001 13:51:54 -0400
>
>There are many applications that could make reasonable 
>use of network RAM.
>Basically everything which need more RAM then currently 
>possible and
>therefore has to swap to disk to process the data 
>(database applications,
>gaussian, etc.). The network latency is high compared 
>to memory latency, but
>not compared to disk slatency. There, network beats it 
>by several orders of
>magnitude.
>I know a lot of applications which have swapfiles of 
>several gigabytes,
>beeing able to keep them in memory (and even if it is 
>network ram) can speed
>then up significantly. And not everybody can afford to 
>buy Itaniums with
>more than 4 Gig's of memory.
>
>Bye, Thommy
>
>-----Original Message-----
>From: Mark Hahn [mailto:hahn at physics.mcmaster.ca] 
>Sent: Thursday, August 16, 2001 9:52 AM
>To: beowulf at beowulf.org
>Subject: Re: Network RAM for Beowulf
>
>
>> >    we as a group of four students are *also* 
>thinking
>> > of implementing Network RAM for a beowulf cluster
>> > (assuming 100Mbps Ethernet ) whereby each node in 
>the
>
>I'm always puzzled why people want to keep trying this:
>have you considered the fairly breathtaking latency of 
>"sharing" pages over a net?  do you really have apps 
>that can 
>tolerate that kind of latency?
>
>
>
>_______________________________________________
>Beowulf mailing list, Beowulf at beowulf.org
>To change your subscription (digest mode or unsubscribe)
> visit
>http://www.beowulf.org/mailman/listinfo/beowulf
>
>--__--__--
>
>Message: 4
>Date: Sat, 25 Aug 2001 13:59:38 -0400 (EDT)
>From: Brian 
>To: "J. M. Pacheco" 
>cc: beowulf at beowulf.org
>Subject: Re: HELP
>
>
>Hi J.M.,
>
>  I ran into this problem as well - the basic problem 
>seems to be that
>linux maps the shared libraries at around the 1GB mark, 
>so you need to
>change this location in the kernel source code.
>
>  Here's a post from late July I sent to the list 
>dealing with the same
>problem.. there's a little more information in that 
>thread, if you look it
>up in the mailing list archives.
>
>  Short answer: Change the source code as below, and 
>you'll be ok.
>
>  Here's more detail:
>----
>Hey guys,
>
>  I think I'm answering my own question now - if anyone 
>else has had this
>problem, here's what SEEMS to be a quick and dirty fix. 
> (Quick and
>dirty being the key words!)  Also, good conscience 
>tells me to write that
>although it's working for me right now, it's not 
>elegant, and who knows
>what problems it may create?
>
>  That said, here goes:
>
>  In the kernel source, in file mmap.c, there is a line 
>that reads:
>
>  addr = TASK_UNMAPPED_BASE;
>
>  .. This is what sets the default address of the 
>shared libs in the
>memory mapping, and it's at 0x40000000 (1G) by default. 
> So change it to,
>for example:
>
>  addr = 0x80000000;
>
>  .. And you should, in theory, have up to ~2GB to use 
>for the codes.  
>People with a little more knowledge than I have can 
>probably shed more
>light on this, and while it's not pretty right now, 
>when I have time I may
>try to work on something nicer.  Right now, I just need 
>these jobs to run.  
>:)
>
>  For more info on this, check out the 
>comp.os.linux.development.system
>newsgroup, back around June 7th or so.
>-----
>
>  Best wishes,
>  - Brian
>
>
>--__--__--
>
>Message: 5
>Date: Sat, 25 Aug 2001 11:36:48 -0700 (PDT)
>From: Ron Chen 
>Subject: Re: HELP
>To: Brian ,
>"J. M. Pacheco" 
>Cc: beowulf at beowulf.org
>
>64-bit Linux should be OK, correct?
>
>-Ron
>
>--- Brian  wrote:
>> 
>> Hi J.M.,
>> 
>>   I ran into this problem as well - the basic
>> problem seems to be that
>> linux maps the shared libraries at around the 1GB
>> mark, so you need to
>> change this location in the kernel source code.
>> 
>>   Here's a post from late July I sent to the list
>> dealing with the same
>> problem.. there's a little more information in that
>> thread, if you look it
>> up in the mailing list archives.
>> 
>>   Short answer: Change the source code as below, and
>> you'll be ok.
>> 
>>   Here's more detail:
>> ----
>> Hey guys,
>> 
>>   I think I'm answering my own question now - if
>> anyone else has had this
>> problem, here's what SEEMS to be a quick and dirty
>> fix.  (Quick and
>> dirty being the key words!)  Also, good conscience
>> tells me to write that
>> although it's working for me right now, it's not
>> elegant, and who knows
>> what problems it may create?
>> 
>>   That said, here goes:
>> 
>>   In the kernel source, in file mmap.c, there is a
>> line that reads:
>> 
>>   addr = TASK_UNMAPPED_BASE;
>> 
>>   .. This is what sets the default address of the
>> shared libs in the
>> memory mapping, and it's at 0x40000000 (1G) by
>> default.  So change it to,
>> for example:
>> 
>>   addr = 0x80000000;
>> 
>>   .. And you should, in theory, have up to ~2GB to
>> use for the codes.  
>> People with a little more knowledge than I have can
>> probably shed more
>> light on this, and while it's not pretty right now,
>> when I have time I may
>> try to work on something nicer.  Right now, I just
>> need these jobs to run.  
>> :)
>> 
>>   For more info on this, check out the
>> comp.os.linux.development.system
>> newsgroup, back around June 7th or so.
>> -----
>> 
>>   Best wishes,
>>   - Brian
>> 
>> 
>> _______________________________________________
>> Beowulf mailing list, Beowulf at beowulf.org
>> To change your subscription (digest mode or
>> unsubscribe) visit
>http://www.beowulf.org/mailman/listinfo/beowulf
>
>
>__________________________________________________
>Do You Yahoo!?
>Make international calls for as low as $.04/minute with 
>Yahoo! Messenger
>http://phonecard.yahoo.com/
>
>--__--__--
>
>Message: 6
>Date: Sat, 25 Aug 2001 22:38:07 +0200
>From: =?iso-8859-1?Q?Jakob_=D8stergaard?= 
>
>To: "J. M. Pacheco" 
>Cc: beowulf at beowulf.org
>Subject: Re: HELP
>"J. M. Pacheco" , 
>beowulf at beowulf.org
>
>On Tue, Aug 14, 2001 at 04:26:54PM +0100, J. M. Pacheco 
>wrote:
>> 
>> 	Hello.
>> 
>...
>> Can anyone help me out of this mess ? 
>> Furthermore, browsing the beowulf mailing list, I got 
>accross some linux
>> limitations in writing huge temperary files. Mine can 
>be 2 GB large. Is
>> that problem solved in kernels 2.4 ? 
>
>The problem resided both in the kernel and standard 
>libraries, so either
>updating glibc *or* updating the kernel won't solve the 
>problem.
>
>However, newer distributions such as RedHat 7.1 have 
>both glibc and kernel
>updated so that this is no longer a problem.
>
>I would think your distribution had this solved too - 
>but I do not know.
>
>The last problem in writing 2+ GB files, is the 
>application (and it's libraries
>including any compiler-included support libraries).   
>If you have code that
>uses a 32-bit signed integer for byte-offsets into your 
>file, well, naturally
>you cannot address more than 2 GB no matter which 
>kernel you're running.
>
>If (and this is *pure* speculation!) g77 generated 
>Fortran code relies on
>a support library for file I/O that uses such 
>addressing, this could be the
>explanation for the problems you see.
>
>Test:  Try dd if=/dev/zero of=test_file bs=8k 
>count=1024k
>
>If you end up with a test_file of size 8 GB, your 
>kernel, glibc, and tools
>(well, dd at least) have proper LFS support.  The 
>problem must then be with
>your code or g77 (or support libraries).
>
>If you do not end up with the large file, I suggest you 
>try another
>distribution that has proper kernel and libraries.   Or,
> see if there are
>updates to your distribution that fix this.
>
>-- 
>........................................................-
>........
>:   jakob at unthought.net   : And I see the elder races,  
>       :
>:.........................: putrid forms of man         
>       :
>:   Jakob Østergaard      : See him rise and claim the 
>earth,  :
>:        OZ9ABN           : his downfall is at hand.    
>       :
>:.........................:............{Konkhra}
>...............:
>
>--__--__--
>
>Message: 7
>Date: Sat, 25 Aug 2001 22:39:59 +0200
>From: =?iso-8859-1?Q?Jakob_=D8stergaard?= 
>
>To: Mark Hahn 
>Cc: beowulf at beowulf.org
>Subject: Re: Network RAM for Beowulf
>Mark Hahn , 
>beowulf at beowulf.org
>
>On Thu, Aug 16, 2001 at 02:52:23PM +0000, Mark Hahn 
>wrote:
>> > >    we as a group of four students are *also* 
>thinking
>> > > of implementing Network RAM for a beowulf cluster
>> > > (assuming 100Mbps Ethernet ) whereby each node in 
>the
>> 
>> I'm always puzzled why people want to keep trying 
>this:
>> have you considered the fairly breathtaking latency 
>of 
>> "sharing" pages over a net?  do you really have apps 
>that can 
>> tolerate that kind of latency?
>
>What latency ?   ;)
>
>I have 100 usec ping latencies on my network.  
>Bandwidth ~ 8-12 MB/sec
>
>I have 8 ms seek latencies on my harddrives. Bandwidth 
>~ 12-16 MB/sec
>
>
>-- 
>........................................................-
>........
>:   jakob at unthought.net   : And I see the elder races,  
>       :
>:.........................: putrid forms of man         
>       :
>:   Jakob Østergaard      : See him rise and claim the 
>earth,  :
>:        OZ9ABN           : his downfall is at hand.    
>       :
>:.........................:............{Konkhra}
>...............:
>
>--__--__--
>
>Message: 8
>Date: Sat, 25 Aug 2001 23:29:38 +0200
>From: =?iso-8859-1?Q?Jakob_=D8stergaard?= 
>
>To: Mark Hahn , 
>beowulf at beowulf.org
>Subject: Re: Network RAM for Beowulf
>Mark Hahn , 
>beowulf at beowulf.org
>
>On Sat, Aug 25, 2001 at 10:39:59PM +0200, Jakob 
>Østergaard wrote:
>> On Thu, Aug 16, 2001 at 02:52:23PM +0000, Mark Hahn 
>wrote:
>> > > >    we as a group of four students are *also* 
>thinking
>> > > > of implementing Network RAM for a beowulf 
>cluster
>> > > > (assuming 100Mbps Ethernet ) whereby each node 
>in the
>> > 
>> > I'm always puzzled why people want to keep trying 
>this:
>> > have you considered the fairly breathtaking latency 
>of 
>> > "sharing" pages over a net?  do you really have 
>apps that can 
>> > tolerate that kind of latency?
>> 
>> What latency ?   ;)
>> 
>> I have 100 usec ping latencies on my network.  
>Bandwidth ~ 8-12 MB/sec
>> 
>> I have 8 ms seek latencies on my harddrives. 
>Bandwidth ~ 12-16 MB/sec
>
>Sorry for following up on my own post, but this needs 
>clarification:
>
>I'm referring to network swap here.  Swapping over fast 
>local networks
>can absolutely make sense.
>
>If you were referring to running separate threads of 
>the same process on
>multiple nodes in a cluster, sharing the same meory 
>space over the network, I
>completely agree with you.  That's not practically 
>possible when the execution
>environment (the operating system kernel) doesn't 
>understand the code
>(compiler-generated native machine-code) it executes.
>
>Even MOSIX migrates complete processes.  For good 
>reason I presume  :)
>
>-- 
>........................................................-
>........
>:   jakob at unthought.net   : And I see the elder races,  
>       :
>:.........................: putrid forms of man         
>       :
>:   Jakob Østergaard      : See him rise and claim the 
>earth,  :
>:        OZ9ABN           : his downfall is at hand.    
>       :
>:.........................:............{Konkhra}
>...............:
>
>--__--__--
>
>Message: 9
>Date: Sat, 25 Aug 2001 19:04:27 -0400 (EDT)
>From: Mark Hahn 
>To: =?iso-8859-1?Q?Jakob_=D8stergaard?= 
>
>cc: beowulf at beowulf.org
>Subject: Re: Network RAM for Beowulf
>
>> > > > >    we as a group of four students are *also* 
>thinking
>> > > > > of implementing Network RAM for a beowulf 
>cluster
>> > > > > (assuming 100Mbps Ethernet ) whereby each 
>node in the
>> > > 
>> > > I'm always puzzled why people want to keep trying 
>this:
>> > > have you considered the fairly breathtaking 
>latency of 
>> > > "sharing" pages over a net?  do you really have 
>apps that can 
>> > > tolerate that kind of latency?
>> > 
>> > What latency ?   ;)
>> > 
>> > I have 100 usec ping latencies on my network.  
>Bandwidth ~ 8-12 MB/sec
>
>I don't think you have 100 us latency for any 
>nontrivial-sized packet.
>my net (FD 100bT, nothing special), shows >.4ms us 
>latency for 1k pings.
>so you'll clearly be into the multiple ms for a 
>useful-sized page cluster,
>say, 32K.
>
>now, I assumed from the original context that someone 
>wanted 
>to do network shared memory.  this is mostly a nutty 
>idea,
>since the granularity is necessarily 4 or 8K, therefore 
>latency
>is nontrivial unless your code is somehow incredibly 
>asynchronous.
>
>network swapping is a somewhat different story, since 
>there's 
>no huge urgency in pushing pages out, and often not 
>that much
>to get them back.  of course, in the former case, 
>you're 
>short of ram, and it's not exactly nice to have to 
>allocate 
>more ram to accomplish the tx, and really you shouldn't 
>free
>the page until you get an ack from the server.  in the 
>latter 
>case, someone will simply sleep until their page 
>arrives, which
>is at least not positive feedback.
>
>
>> > I have 8 ms seek latencies on my harddrives. 
>Bandwidth ~ 12-16 MB/sec
>
>that's pretty miserable bandwidth.  it's basically 
>impossible to buy a modern
>IDE disk, for instance, that sustains less than 20 MB/s 
>on inner/slow tracks,
>and most peak close to 40 MB/s.  not to mention how 
>easy it is to 
>stripe them, or just throw 15krpm scsi at the problem.
>
>> I'm referring to network swap here.  Swapping over 
>fast local networks
>> can absolutely make sense.
>
>networks/nics suck, relative even to ide controllers.  
>there's just
>no getting around that.  it's routine to launch off a 
>128K scatter-gather,
>busmaster command to a cheap UDMA controller - yes, 
>read latency can be 
>unpleasant, but the bandwidth is great, and the 
>overhead is minimal.
>
>you simply can't say that about networks, which is a 
>shame.  the world
>would be a far better place if I could at least send 
>8-9K jumbos,
>preferably with a sane, zero-copy-friendly s-g 
>interface.
>
>yes, gigE helps, as do smarter nics, zcopy, jumbograms, 
>even exotica
>like STP.
>
>but have you looked at how much CPU is eaten by a gigE 
>card 
>streaming, say, 50 MB/s, versus a cheap dual-channel 
>IDE doing so?
>
>in summary: swapping or shmem over a net is attractive 
>at the surface,
>but crunch the numbers and you find it's only a win in 
>very special cases.
>(that said, I'll go back to tuning up my 112-cpu SC40 
>that has ~200 MB/s
>very smart interconnect ;)
>
>regards, mark hahn.
>
>
>--__--__--
>
>Message: 10
>Date: Sat, 25 Aug 2001 19:23:27 -0400 (EDT)
>From: Mark Hahn 
>To: beowulf at beowulf.org
>Subject: Re: HELP
>
>>   I ran into this problem as well - the basic problem 
>seems to be that
>> linux maps the shared libraries at around the 1GB 
>mark, so you need to
>> change this location in the kernel source code.
>
>more generally, mmaps start at 1G by default (not just 
>shlibs).
>it's pretty cool that you can just "cat 
>/proc/self/maps" to see this.
>
>and you can indeed hack the kernel to change the 1G; 
>there's actually 
>a somewhat more involved patch to make the mmap arena 
>grow *down*,
>so you can fix the max size of the stack instead of the 
>heap.
>it's conceivable this might get into 2.5.
>
>it's also true that you can use a modern glibc whose 
>malloc 
>will use both heap and mmap arenas; that doesn't give 
>you a 
>big contig allocation, though.
>
>you can also avoid the problem *entirely* by avoiding 
>mmaps!
>I posted a proof-of-concept of this here or linux-kernel
>sometime in the past year - just avoid using stdio and 
>any dynlibs ;)
>(stdio seems to want to mmap a single page for an 
>internal buffer,
>even if you statically link...)
>
>>   addr = TASK_UNMAPPED_BASE;
>
>note also that some Linux ports have this configurable -
>I think it's ia64 that has it set in /proc...
>otoh, it's sort of moot on big-address hardware ;)
>
>regards, mark hahn.
>
>
>--__--__--
>
>Message: 11
>Date: Sun, 26 Aug 2001 13:21:59 +1000 (EST)
>From: Kim Branson 
>To: Tom Poe 
>cc: "Eric T. Miller" ,
>"Lambe, Dave" ,
>"Beowulf (E-mail)" 
>Subject: Re: Need to do something Useful
>
>
>
>
>the code (some of it is based on work done at the UCSF, 
>its a minimal
>license for academic work type of deal... industry 
>types pay big $$ ) So
>there might be legal hassles for non academic work. 
>
>It sure like to set it up as a beowulf.org project, but 
>i would need some
>help...volunteers?
>
>kim
>
>________________________________________________________-
>______________ 
>
>Mr Kim Branson
>Phd Student
>Structural Biology
>Walter and Eliza Hall Institute
>Royal Parade, Parkville, Melbourne, Victoria
>Ph 61 03 9662 7136
>Email kbranson at wehi.edu.au
>
>________________________________________________________-
>______________ 
>
>
>On Wed, 15 Aug 2001, Tom Poe wrote:
>
>> Hi, again:  Forgot to mention, this looks to me like 
>it's a really neat
>> beowulf.org project, or HA.org project.  What do you 
>think?  I'd like to do
>> a bid, but maybe it's something the Open Source 
>community can rally around.
>> Who knows?  Thanks, Tom
>> ----- Original Message -----
>> From: Kim Branson 
>> To: Eric T. Miller 
>> Cc: Lambe, Dave ; Beowulf 
>(E-mail)
>> 
>> Sent: Wednesday, August 15, 2001 4:18 PM
>> Subject: RE: Need to do something Useful
>> 
>> 
>> >
>> >
>> > Of you like you could install globus on your 
>machine and we can add it to
>> > our global grid for drug design applications.
>> >
>> > see http://www.csse.monash.edu.au/~rajkumar/vlab/ind-
>ex.html
>> > for more details.
>> >
>> > We are interested in designing small molecule 
>atagonists for malaria and
>> > leishmania.
>> >
>> > failing that theres always seti or something....
>> >
>> > kim
>> > ____________________________________________________-
>__________________
>> >
>> > Kim Branson
>> > Structural Biology
>> > Walter and Eliza Hall Institute
>> > Royal Parade, Parkville, Melbourne, Victoria
>> > Ph 61 03 9662 7136
>> > Email kbranson at wehi.edu.au
>> >
>> > ____________________________________________________-
>__________________
>> >
>> >
>> > On Wed, 15 Aug 2001, Eric T. Miller wrote:
>> >
>> > >
>> > > >About 2 months ago, I setup a small cluster 
>using Scyld. The
>> > > >developers/programmers have been keyholed into 
>another project for the
>> time
>> > > >being. I would like to do something with the 
>cluster as it's just using
>> > > >electricity & creating heat.
>> > > >Is there a way to fire up dnetc (or similar) on 
>all the nodes? I
>> apologize
>> > > >for my lack of *nix knowledge (I'm a 
>hardware/setup guy). TIA
>> > > >
>> > > >
>> > > >Dave
>> > >
>> > >
>> > > Yes! I had a similar question about a week ago 
>that didn't get much
>> > > response.  I too am new to clusters, and I just 
>want to do something
>> useful
>> > > and interesting with my new creation. It is 
>currently just a very
>> > > intelligent space heater.
>> > >
>> > >
>> > > _______________________________________________
>> > > Beowulf mailing list, Beowulf at beowulf.org
>> > > To change your subscription (digest mode or 
>unsubscribe) visit
>> http://www.beowulf.org/mailman/listinfo/beowulf
>> > >
>> >
>> >
>> > _______________________________________________
>> > Beowulf mailing list, Beowulf at beowulf.org
>> > To change your subscription (digest mode or 
>unsubscribe) visit
>> http://www.beowulf.org/mailman/listinfo/beowulf
>> >
>> 
>> 
>> 
>> _______________________________________________
>> Beowulf mailing list, Beowulf at beowulf.org
>> To change your subscription (digest mode or 
>unsubscribe) visit http://www.beowulf.org/mailman/listin-
>fo/beowulf
>> 
>
>
>--__--__--
>
>Message: 12
>Date: Sun, 26 Aug 2001 12:50:56 +0200
>From: "andreas boklund" 
>To: 
>Subject: Re: Network RAM for Beowulf
>
>Just a small
>If you take a look at http://www.mosix.org/ they have a 
>different approach to Network RAM. Trying to move the 
>process to the Data instead of moving pages over the 
>network. Interesting project, i want to know where they 
>end up...
>
>Best Regards
>//Andreas
>
>
>--__--__--
>
>Message: 13
>charset="iso-8859-1"
>From: Eray Ozkural (exa) 
>Organization: Bilkent University
>To: Thomas R Boehme ,
>"'Mark Hahn'" , 
>beowulf at beowulf.org
>Subject: Re: Network RAM for Beowulf
>Date: Sun, 26 Aug 2001 17:31:06 +0300
>
>-----BEGIN PGP SIGNED MESSAGE-----
>Hash: SHA1
>
>On Saturday 25 August 2001 08:51 pm, Thomas R Boehme 
>wrote:
>> There are many applications that could make 
>reasonable use of network RAM.
>> Basically everything which need more RAM then 
>currently possible and
>> therefore has to swap to disk to process the data 
>(database applications,
>> gaussian, etc.). The network latency is high compared 
>to memory latency,
>> but not compared to disk slatency. There, network 
>beats it by several
>> orders of magnitude.
>> I know a lot of applications which have swapfiles of 
>several gigabytes,
>> beeing able to keep them in memory (and even if it is 
>network ram) can
>> speed then up significantly. And not everybody can 
>afford to buy Itaniums
>> with more than 4 Gig's of memory.
>
>I don't think you could say that in general. The 
>network is fast for a few 
>large messages perhaps but not for many small ones. The 
>problem, IMHO, stems 
>from the fact that there is no low level way to 
>transform a serial program to 
>a parallel one! The more non-local memory access 
>pattern in the original
>application, the worse speed-up you will get. You would 
>get good speedup, it
>seems to me, only if the problem is embarrasingly 
>parallel. By speedup here I
>mean the ratio of running time to single node/virtual 
>memory. That is, a 
>distributed shared memory system -no matter how 
>efficient or optimal- is not 
>likely to provide a scalable way to increase memory 
>available for 
>applications, especially if you want to utilize more 
>than a single node's CPU 
>for large tasks.
>
>I think that is also the same reason why you can use 
>virtual memory for
>running some algorithms while it would be infeasible 
>for others.
>Theoretically speaking, an out-of-core algorithm can be 
>shown to be
>equivalent to a certain parallel architecture.
>
>That is not to say that DSM is a cursed enterprise, but 
>rather that such
>methods have to be complemented by parallel programming 
>support.
>
>Regards,
>
>- -- 
>Eray Ozkural (exa) 
>Comp. Sci. Dept., Bilkent University, Ankara
>www: http://www.cs.bilkent.edu.tr/~erayo
>GPG public key fingerprint: 360C 852F 88B0 A745 F31B  
>EA0F 7C07 AE16 874D 539C
>-----BEGIN PGP SIGNATURE-----
>Version: GnuPG v1.0.6 (GNU/Linux)
>Comment: For info see http://www.gnupg.org
>
>iD8DBQE7iQgqfAeuFodNU5wRAmDdAJ9fjAHeh+DZojIEMEW/jbuKnM9b-
>GACfdTP/
>L0bff8QqLzx3XqLL0B6drgM=
>=LJLN
>-----END PGP SIGNATURE-----
>
>
>
>--__--__--
>
>_______________________________________________
>Beowulf mailing list
>Beowulf at beowulf.org
>http://www.beowulf.org/mailman/listinfo/beowulf
>
>
>End of Beowulf Digest


 







More information about the Beowulf mailing list