What method is best (or is there any...), if I want to read table tiitm001 by using several spindles at the same time? (think about bdbpre6.1 or full table scan, where rows are read one by one)
Let's say, I want separate fields to different tablespaces, which are in different disks in Oracle. How it's possible?

hpux11 BaanIV Oracle 8.1.6

Is hardware/software striping an option?

I'm just thinking different possibilities. Let's start from an idea, that there is only a few very I/O-intensive tables like itm001, sls,pur and few big gld-tables. I'd like to start performance tuning from these tables and not going to stripe whole database with LVM extent based striping (hpux mirrordisk).
I have read tens of documents during few last days about striping and I'm still not sure is there any quarantee that read performance will get any better. Of course it helps with I/O balancing, but my goal is to get more read performace compared to mirrored pair (SC10).

It is fairly straightforward to place discrete tables on their own tablespace to avoid contention or to better balance IO.

The SC10 is not the best performing disk option out there so if you can stripe the IO using software RAID 0+1 and then use a degree of parallelism by increasing the number of IO slaves or async IO, then you will definitely get improved performance.

LVM manual striping (or chunking) is the manual placement of blocks by selectly using lvextend with explicit disk selection. This can improve IO, but is inconsistent when compared to the previous method.

-- Jim