Beefy Boxes and Bandwidth Generously Provided by pair Networks
Pathologically Eclectic Rubbish Lister
 
PerlMonks  

Muy Large File

by BuddhaLovesPerl (Sexton)
on Mar 14, 2005 at 07:24 UTC ( #439181=perlquestion: print w/replies, xml ) Need Help??
BuddhaLovesPerl has asked for the wisdom of the Perl Monks concerning the following question:

Venerable greetings to the monks,

I have a single file that ranges daily from 45-50Gig on a Solaris 8 server with 16G ram and 8 900Mhz cpu's. The task is simply to replace specific control characters with a space. Due to space limitations, this transform must be conducted in-place on the original file. The records are fixed and ascii.

Using perl with a single loop/unpack and s///, this is taking over 4 hours. I am not sure how much longer this would take to complete as I killed it. The transform rate at kill time was about 400M per hour.

The sys admin wraith types don't want dev orc types running 24h (or more) for each file. Ergo, I need to find the fastest method possible to accomplish this.

2 questions if you please.

1) Without sounding like a cliche, what is the 'fastest' way to get this done? Is a combination of Sys::Mmap and forks the fastest ticket out of this hell?

2) With all the icing, could this ever be done in less than 12 hours?

A lagging japh,
--Paul

Replies are listed 'Best First'.
Re: Muy Large File
by BrowserUk (Pope) on Mar 14, 2005 at 09:01 UTC
    ... this is taking over 4 hours....

    You're doing something wrong :).

    The following shows Perl processing a 32 GB file in-place, finding and replacing 30% of it's contents in under 25 minutes; on a single cpu 512 MB ram machine. (the process only uses 3 MB of ram).

    #! perl -slw use strict; our $BUFSIZE ||= 2**20; open my $fh, '+< :raw', $ARGV[ 0 ] or die $!; while( sysread $fh, $_, $BUFSIZE ) { tr[123][123]; sysseek $fh, -length(), 1; ## Updated per Dave_the_m's correction +below++ syswrite $fh, $_; } close $fh; __DATA__ [ 8:31:52.64] P:\test>439181 data\integers.dat [ 8:54:43.92] P:\test>dir data\integers.dat Volume in drive P has no label. Volume Serial Number is BCCA-B4CC Directory of P:\test\data 14/03/2005 08:54 34,359,738,368 integers.dat 1 File(s) 34,359,738,368 bytes

    Examine what is said, not who speaks.
    Silence betokens consent.
    Love the truth but pardon error.
    Lingua non convalesco, consenesco et abolesco.
      One slight nit: if the file size isn't a multiple of the buffer size, the final seek will seek back too far and corrupt the final block.
      while(sysread $fh, $_, $BUFSIZE ) { tr[123][123]; sysseek $fh, -length(), 1; syswrite $fh, $_; }
      (untested).

      Dave.

      Another problem is with data that crosses the buffer.
      -- gam3
      A picture is worth a thousand words, but takes 200K.
Re: Muy Large File
by perlfan (Curate) on Mar 14, 2005 at 19:24 UTC
    I have a single file that ranges daily from 45-50Gig on a Solaris 8 server with 16G ram and 8 900Mhz cpu's.

    How can you not be taking advantage of that horsepower?

    I would seriously look into MPICH's implementation ROMIO http://www-unix.mcs.anl.gov/romio/ (MPI Standard 2.0) http://www.mpi-forum.org/docs/mpi-20-html/node171.htm#Node171

    If it has to be Perl, then I would certainly look into parallelizing this application - as a brute force approach, split the file 8 ways, run a process to take of each piece, then join the darn things back together. Even with the splitting and rejoining, I am sure it would be faster than what is happening right now.
      Since he needs it to be in-place, I suggest 8 processes that work on the same file at the same time... if that's possible on Solaris. YOu can use seek for that.
Re: Muy Large File
by TilRMan (Friar) on Mar 14, 2005 at 10:33 UTC

    Update: As thor points out, my solution here is quite wrong. Sorry.

    On unixy platforms, Perl has exactly what you need already built in. (Windows users may need a binmode somewhere.)

    #!/usr/bin/perl -w -pi use strict; tr/A-Z/ /;

    If your records do not have newlines at the end, then you will need to set the record length. Add the following line to the script, replacing 4096 with the record length:

    BEGIN { $/ = \4096 }

    The magic is in the -pi which turns on in-place editing in a loop (more at perlrun). Then the tr/// operator runs on every record, replacing the offending characters with spaces (more at perlop). Use tr/// instead of s///; it's probably faster and safer.

    Note: Code is mostly untested. Use with caution.

      The problem is that while the -i switch is the "in place" switch, it isn't really in place. From perldoc perlrun
      It does this by renaming the input file, opening the output file by the original name, and selecting that output file as the default for print() statements.
      As per the OP, the file is too large to do this with.

      thor

      Feel the white light, the light within
      Be your own disciple, fan the sparks of will
      For all of us waiting, your kingdom will come

Re: Muy Large File
by Anonymous Monk on Mar 15, 2005 at 20:27 UTC
    Are other processes running at the same time as you do your conversion? If there's a set of constant-HD-use programs running, the HD will have to seek for many of the block read/writes, and it's builtin cache won't be anywhere near as effective. So - my first guess? Try shutting down all services and kicking all users, if that's amenable to the wraith types. Run it for an hour and see how far through it gets. Even if this isn't the case, be cautious with parallel processes; HD misses are on the order of milliseconds iirc, which means that at some number of processes you're going to have the bottleneck come from HD access not from CPU time and RAM accesses. To check: use a clock-tick timer and time a read of one (non-cached, and make sure it seeks!) block off the HD. Ditto for a write (they should be almost the same, though if it caches the write - eg. doesn't use write-through - it could be longer). Then time a search/replace. Betcha the latter is faster.
      Wow. Many deep bows of reverence for all that responded. As UK inferred, I was (indeed) doing something wrong. Based on the above suggestions, this was the script tested:

      #!/usr/local/perl5.6.1/bin/perl -slw
      use strict;
      our $BUFSIZE ||= 2**30;
      open my $fhi, '+<', "/data/p_dm200/ndm_ip_pull/test_customer1" or die $!;

      while( sysread $fhi, $_, $BUFSIZE ) {
      tr^M ;
      sysseek $fhi, -length(), 1;
      syswrite $fhi, $_, $BUFSIZE;
      }
      close $fhi;

      which was tested against an 8,595,447,728 byte file. The time output was:
      real 10m5.95s
      user 1m48.55s
      sys 0m17.24s

      An amazing 10 minutes. I checked the output and it looks exactly as expected. I even retested 3 times and each time the results were similar.

      Ok, now I am getting greedy and curious as to if this can be optimized more?? I ran top during this session and saw that SIZE and RES were both around 1026M throughout the duration and only 1 cpu seemed used. Would increasing BUFSIZE help performance linearly? If I was capable (and I am not) would either shared memory threads or parallel forks produce big gains? Any other low-hanging fruit?

      Perlfan, the ROMIO seemed interesting but I could not find a perl sample. Still it seemed interesting. Anonymous Monk, please forgive my ignorance but what does HD mean?

      A sincere thanks to all,
      --Paul

        Using a larger buffer size may increase throughput slightly, but then it may not. It will depend upon many factors mostly to do with your file system buffering, disk subsystems etc. The easy answer, given it's only taking 10 minutes is to try it.

        As far as using threads to distribute the load across your processors is concerned, it certainly could be done, and could in theory, give you near linear reductions per extra processor.

        But, and it's big 'but', how good is your runtime library's handling of multi-threaded IO to a single file?

        On my system, even using sys* IO calls and careful locking to ensure that only one thread can seek&read or seek&write at a time, something, somewhere is getting confused and the file is corrupted. I suspect that even after a syswrite completes, the data is not yet fully flushed to disk before the next seek&write cycle starts.

        So, maybe you could get it to work on your system, but I haven't succeeded on mine, and I am not yet entirely sure whether the problem lies within Perl, the OS, or some combination of the two.

        If you feel like trying this, please come back and report your findings. If you need a starting point, /msg me and I can let you have my failing code.


        Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.
        Lingua non convalesco, consenesco et abolesco.
        Rule 1 has a caveat! -- Who broke the cabal?

        By HD anonymonk means Hard Disk. The seek times to move the heads around a hard drive are slow compared to memory access and geological compared to processor cache. What this means is processes that are dedicated to doing something to a file are normal disk IO bound. Lets not mention network latencies for now.

        If you do manage to split this into threads you may actually reduce performance as each time a different thread gets a shot at it, it forces the HD to drag it's heads over to a completely different part of disk. A single thread reading the file sequentially will not be making the heads seek so much, assuming the file is not desperately fragmented on the media.

        Then there are other users competing for those heads and tasking them off to the boondocks of the drive as far as your data is concerned which is why it was suggested you kick the lusers to try and get the disk all to yourself.

        Cheers,
        R.

        Pereant, qui ante nos nostra dixerunt!

Log In?
Username:
Password:

What's my password?
Create A New User
Node Status?
node history
Node Type: perlquestion [id://439181]
Approved by Corion
Front-paged by Courage
help
Chatterbox?
and all is quiet...

How do I use this? | Other CB clients
Other Users?
Others pondering the Monastery: (6)
As of 2018-05-27 14:53 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?
    Notices?