Acorn Fragmentation Avoider

Description:
   VMS disks tend to become fragmented rapidly once placed in use,
causing slower file access than might be possible, and making more
work for defragmenting utilities later. This happens mainly for two
reasons:
   First, whenever a file is deleted or truncated in VMS, the blocks
are placed in an "extent cache" which is then used first in allocating
new space. Unfortunately, this space tends to be in tiny, scattered
chunks and guarantees that newly allocated space is badly fragmented.
   Second, many utilities write long files, particularly journal files,
and use small extents (~5 blocks is common) even though the files may
become thousands of blocks long. Because these files are held open
most of the time, they not only fragment, but are difficult to defragment
because defragmentation of open files is far more difficult and intrusive
than of files which are closed. As these files pepper a disk with little
bits of themselves, they make defragmentation of the rest less effective.

The Fragmentation Avoider consists of an intercept driver and a
control utility which reduce the rate of VMS disk fragmentation
in two ways:

1. Files are extended in "contiguous best try" fashion, preventing
	the accumulation of many tiny extents in extent cache which
	would otherwise cause new files to be allocated in small
	bits. 
2. Files are extended by a fraction of their current size when this is
	larger than the default, subject to minimum and maximum
	extents and to a constraint of never forcing the extent
	to be more than 1/8 of volume free space. Thus files tend
	to be extended less often and in larger pieces than by default
	which leads to their being far less fragmented than otherwise.

These actions are configurable on a per volume basis, and FragmentAvoid
can be enabled or disabled separately on any disk on a system. Configuration
options include:

*  Selection of the fraction to extend a file by (default is 1/4 of
	current size, but can be tailored).
*  Selection of maximum number of blocks to extend by, even for giant
	files
*  Selection of minimum number of blocks to extend by
*  Selection of how often to force contiguous best try extension. This
	is normally done every time, but can be done every Nth time
	where N can range from 1 to 1000. (Note that requests for
	contiguous extension are not touched.)
*  Selection of whether to modify extension amounts where the application
	has selected using volume default extension only, or whether
	to do this every time. (Using the extension modification every
	time overcomes problems with utilities which write very long
	files with hardcoded extensions of only a few blocks, leading
	to badly fragmented files. Occasionally, though, this leads to
	the need to manually truncate files via the $SET FILE/TRUNCATE
	command. If size is altered only for users of volume default
	allocation, files never need to be truncated, but some files
	will fragment the disk.)

   Fragmentation Avoider works on all VMS disks and is a "set it and
forget it" utility. Users will notice it in that long files will
tend to be written faster (some measurements show up to a 30% speedup)
due to less frequent use of the (expensive) extend operation. They will
also notice slower volume fragmentation if they monitor it, and will
generally see fewer extents in any file if they use the DUMP/HEADER
command to examine this. The effect is particularly dramatic on system
disks with files like ACCOUNTNG.DAT or the OPCOM logfiles, which are
generally hard to defragment.


   A menu-driven setup script makes setup of FA simple to do, even
without training.

Processors supported: All VMS processors, running VMS 5.5 and above on
	VAX, or VMS 6.1 or later on Alpha. (VMS 1.5 support on Alpha
	works if you reassemble; the source code is here.)

Requirements:
	Sufficient memory to hold one multiunit driver in pool plus
	its control blocks. No other software is required.