I have been trying and testing Linux (Slackware 10.2) for the past 2 weeks and now is time to pose some questions, first a bit of history as to why I arrived here.

We produce a piece of DOS Software that proves very useful for many clients; this software has migrated over many years through many OS’s. The software is multi-user and has run on CDOS, Real32, LanTastic, Novell and into Windows Networking, it is with the latter that we are having problems particularly with Win2k3 servers where the filing system appears to be very unstable. It was about a month ago a business associate suggested to me that I might want to look a Linux as an alternative Server OS.

With this in mind I have downloaded Slackware and built a test bed server to evaluate the OS, I have installed the OS on a old PIII 800 with 512Mb RAM and 80Gb IDE HDD, I have kept the OS to the minimum I require (No KDE or X-Windows), initial testing appears to be good and I have made the following observations.

1. If on a workstation I run a file rebuild on a database stored on a W2k3 server the rebuild takes about 12Mins, whereas doing the same task on the same database stored on the Linux box takes a little over 2Mins – any ideas Why?

2. If while running the rebuild I crash the application, the Linux box appears to close the files that the user had open – is that the expected result of a user dropping out!

3. Is the OS’s File handling more like that of older OS’s?

4. Bearing in mind that I intend to run the box as a Server, is Slackware a good choice or is there another OS for Server Use?

Well that’s about it any feed back would be appreciated!

Regards

Chris

3 -- what do you mean?

4 -- 100%. Check out the Slackware forum here at LQ for a plethora of questions and answers to many Slack-related elements

hi,
Linux kernel provides a Virtual File System that many other filesystems can interface with and that can interface with other filesystems. generally it's
superblock
inode
file
dentry

and on top a filesystem like reiser or ext3 that journals or logs metadata i-nodes, free block allocation maps, i-nodes maps, etc operations in a log file as they occure for consistency and stability much like a database would.

3. i guess if you mean like older unixes the answer is yes only of course structures have been enlarged to function in a larger disk environment and with better, more apropriate, addressing algorithms and what not.

2. i imagine the connection is run as a seperate "forked" process so yes all open file descriptors would be closed.

1. possibly Linux use of cached mamory pages and cached text and file system info buffers. ?? we don't know how windows works .

As for 1., it is prob at least due in part to the fact that (as you said), you can build Linux to only have software you want ie it's fine-grained. MSWin is more monolithic.
There are always arguments about benchmarking the 2 OS's, but the instinctive feel to me at least is that Linux seems to be faster. Certainly you can run a decent system on a lower spec piece of HW for the same result.
As it happens, a mate of mine is working at a company and recently designed & built a system using Apache/MySql/PHP, originally on Windows. He's never used Linux before, but decided to try it for various reasons. His (un-prompted) comment was that it's definitely faster after moving it to Linux (in his case RH FC3 or 4).
YMMV
Incidentally, this is quite amusing http://os.newsforge.com/article.pl?s...30222&from=rss

Yes, it was rather amusing. Thanks for the link!