Disk Defragmenters for Windows Alternatives

Posted on
Share on Google+Share on LinkedInShare on FacebookShare on RedditTweet about this on TwitterEmail this to someone

Insofar as non-Microsoft operating systems, NetWare makes efficient use of elevator seeking, a technique that optimizes head movement by gathering related requests, waiting a defined interval, reordering those requests and then making clean sweeps across the disks. This limits read/write head movement. See TIDs 2912014and 10027705. Set parameters that control disk access include:

 

 

  • Maximum Concurrent Disk Cache Writes
  • Maximum Concurrent Directory Cache Writes
  • NCP File Commit
  • Dirty Disk Cache Delay Time
  • Dirty Directory Cache Delay Time
  • Read Ahead LRU Sitting Time Threshold

 

Use iMonitor to check statistics on Concurrent Disk Requests and Dirty Cache buffers to assist in tweaking these parameters. This and other tuning parameters are explained in TID 10012765. Fragmentation is not usually as much of a factor in NetWare versus other OSs until the disk is nearly full. Heavy disk usage servers, as well as loaded servers, will benefit from defragmenting, though. Portlock Software (www.portlocksoftware.com/products/defrag/) makes a Novell defragmenter and has a Linux version in beta.

Though some Web sites state emphatically that Linux doesn’t fragment, that is a myth. A paper written by the developers of the Ext2fs (e2fsprogs.sourceforge.net/ext2intro.html#section:ext2fs), a common Linux file system, explains how fast symbolic links are implemented. Briefly, among other optimizations, the target name is stored in the inode, not in a data block, and block groups are used to cluster together related inodes and data blocks. A directory, really just a list of entries, is stored in memory as follows: An inode (Index node) contains information on UNIX data structures. An inode table is stored in memory and contains a list of numbers (inode numbers), paired directory file names, file permissions and type, time stamps, size, etc. Searches are done on the inode number, after the pairing is done on the first request. A clearer, graphical description of how data is stored (using BSD) is at www.dysphoria.demon.co.uk/OperatingSystems1/5_file_allocation_unix.html. So fragmentation does occur, but again, head movement is minimized, and common UNIX file systems allow keeping track of millions of blocks making up a file—hence, fragmentation.

Doug Mechaber is a network engineer and architect who enjoys finding new ways to make his work easier. Douglas has written for Network Computing and remains active in several local user groups. He now works as an independent consultant. Send him the names of your favorite utilities at certmag@att.net.

 

Share on Google+Share on LinkedInShare on FacebookShare on RedditTweet about this on TwitterEmail this to someone
cmadmin

ABOUT THE AUTHOR

Posted in Archive|

Comment:

Leave a comment

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>