RE: stat(), st_size and st_allocate

Bruce:

midrange-l-request@midrange.com wrote:

>  13. stat(), st_size and st_allocate (Bruce Vining)
>
>st_allocsize represents the number of bytes associated with the file by the
>owning file system (in the case of your QNTC usage, Windows 2000).  This
>would include data, headers, reserved growth space, and pretty much
>anything else the file system associates with the file.

This fits just about everything that was assumed. Seems reasonable.

>  In the case of
>stat() returning 69632 (68K), what does Windows show as being the allocated
>size?  Some quick testing here showed Windows returning the same size as
>stat() on the iSeries.

I almost wish I had even the rudimentary tools that this requires. The simple 
response is that I'm more or less stuck with running stat() over /QNTC. I am by 
no means a Windows coder... VB, C, C++, C# or otherwise.

>  Someone also wondered if you perhaps had RAID
>protection in effect.  This may (may being a disclaimer here...) be causing
>a larger allocation than what you may expect (like maybe (disclaimer again)
>64K for data and 4K for header).  This value can be influenced by you (at
>least for Root, QOpensys, and user-defined file systems) by way of the
>QP0L_ATTR_DISK_STG_OPT attribute of Qp0lSetAttr().

Probably irrelevant for my purposes, so no disclaimer needed. Remote Win2K 
Server systems are the critical ones at the moment. Local files are relatively 
easy.

>st_blocksize represents the preferred size for an IO operation by the file
>system, and is not currently used by iSeries (but is set and returned as
>part of the file system implementation).

>If the desire is to manage by object size/space, I would vote for using
>st_allocate.

I've decided to check both. I'll use st_size as my absolute threshold, and 
st_allocsize will be secondary. st_size will be tested against something like 
thresholdsize(filen) and st_allocsize against thresholdsize(filen)+10% or 
similar. If either threshold is passed, I'll trigger the procedure.

I ran the tests again with a 10MB threshold to see how the st_size/st_allocsize 
ratio progressed. At 10MB, Win2K Server reported a much better ratio than 
OS/400. That may be because OS/400 begins allocating larger spaces as the 
object size continues to grow or because Win2K simply handles streamfile 
objects better.

In any case, it certainly seems that _anyone_ managing space via the IFS APIs 
should pay attention to st_allocsize.


>I'll see if we can get a little more description on these attributes in the
>next publishing of the API.

Now that a couple real-world discussions have happened, it'll probably be 
easier to prepare useful descriptions. They'll certainly be appreciated.

Thanks for contributing.

Tom Liotta

--
Tom Liotta
The PowerTech Group, Inc.
19426 68th Avenue South
Kent, WA 98032
Phone  253-872-7788 x313
Fax    253-872-7904
http://www.powertechgroup.com


__________________________________________________________________
The NEW Netscape 7.0 browser is now available. Upgrade now! 
http://channels.netscape.com/ns/browsers/download.jsp

Get your own FREE, personal Netscape Mail account today at 
http://webmail.netscape.com/