Hi Frank,
FR> Netware performance actually increases because of fragmentation.
FR> It's because of the elevator seeking technology that no other
FR> operating system has.
I understand elevator seeking but I fail to see how physical
fragmention improves performance rather than the opposite.
Correct me where I am wrong. Elevator seeking is the process of
storing read/write requests to maximize the time the drive heads
are in any given area. This presumes contiguous file blocks to
_some_ degree. The very nature of the varying size of files being
written and deleted guarantess the phenomenon known as fragmentation.
To be more detailed, the FAT ( collection of info about every block
within the volume ) is cached in server RAM.
"Files in Netware are chains of FAT entries. That is, a large file
consists of a series of FAT entries that are linked together, in
sequential order, to form a chain. Although the actual data of the
file is located on volume blocks that resolve to physical disk
sectors, the FAT provides information to the Netware file system that
enables the file system to read or write the file into or out of
memory from the correct physical disk sectors, and in the correct
order." - Netware 4 by Sunil Padiyar.
The fact remains the data must be read from physical disk sectors
which may not be contiguous. Even with the FAT being cached, and
elevator seeking, the physical action of head movement has to happen
when data is not contiguous. We may be talking only a micro-second
per dozen requests but multiplied by the number that is done in a day
can mean a lot more work than is necessary by the drive heads.
Of what use, if any, is compression in this discussion of performance
at the disk level? Unless Netware is doing physical disk sector
management at some level, performance deteioration is inevitable. An
abundance of RAM and elevator seeking no doubt minimizes this to a
very high degree but my noodle fails to grasp how disk fragmentation
in Netware does not exist. Thus: If it exists there is a cost to
performance.
If disk fragmentation is an issue it will manifest itself more as
time goes on. Certainly this issue can be put to rest if anyone has
made any performance measurements of a server before total backup and
then restoring it. The process of tape backup is one of queuing all
file fragments together in the right order to form a contiguous write
on tape; a total restore will automatically create a 100% contiguous
file structure on the disk.
Have fun with it!
Regards [ or not :-))) ]
Eric
* PMDBM V1.75b #14 *
--- PCBoard (R) v15.23/M 5
---------------
* Origin: FidoNet þ IDS Place þ 416-252-1133 33.6Kbs FTS/QWK (1:259/222)
|