When I time sed for this (using a ~250K line log file as input), I get: time sed -n 10000,13000p mysqld.log > /tmp/file1
real 0m0.220s user 0m0.156s sys 0m0.044s When I use tail and head, I get: time ( tail -n+10000 mysqld.log | head -n 3001 > /tmp/file2 ) real 0m0.014s user 0m0.004s sys 0m0.010s So at least as far as CPU usage is concerned, I'd go with head and tail. Using sed may have more overhead because it's prime function is not to split files in this way. - Richard Trey Darley wrote: > Say you've got a simple ascii text file, say, 250,000 lines long. Let's > say it's a logfile. Suppose that you wanted to access an arbitrary range > of lines, say, lines 10,000 - 13,000. One way of doing this is: > > <snip> > sed -n 10000,13000p foobar.txt > </snip> > > Trouble is, the target systems I need to exec this on are ancient and > don't take very kindly to the io hammering this delivers. Can you suggest > a better way of achieving this? > > As these *are* logs I'm dealing with I have already implemented rotation > frequency. That helps, but I'm still facing performance issues. I'm > specifically looking for input as to whether my sed-based approach can be > improved. > > Because of extremely limited network bandwidth pulling the files off the > wire and processing them on more studly hardware isn't an option. Also, I > cannot install any binaries on these remote systems. Standard POSIX > toolkit is all I've got to work with. :-/ > > Many thanks, y'all! > --Trey > ++----------------------------------------------------------------------------++ > Trey Darley - Brussels > mobile: +32/494.766.080 > ++----------------------------------------------------------------------------++ > Quis custodiet ipsos custodes? > ++----------------------------------------------------------------------------++ > > > > _______________________________________________ > Discuss mailing list > Discuss@lopsa.org > http://lopsa.org/cgi-bin/mailman/listinfo/discuss > This list provided by the League of Professional System Administrators > http://lopsa.org/ > _______________________________________________ Discuss mailing list Discuss@lopsa.org http://lopsa.org/cgi-bin/mailman/listinfo/discuss This list provided by the League of Professional System Administrators http://lopsa.org/
