Please:
- describe what caused this behavior (from 6 to 21 hours)
- post "onstat -d" output
- post "onstat -l" output
- do not use fuzzy checkpoints
- do not use RAID5 (see www.baarf.com and read carefully)
- if you have 64bit fedora, use 64bit informix
- if you can, use v11.50 (64 bit) - you can set it to do autotuning
- in v11 there is recommendation for much bigger physical log (your is only 2000kB and you can make it much bigger - e.g. 2000000kB)
- your SHMVIRTSIZE is to big for your physical memory (you configured 5000000 kB = 4.76GIG which does not fit in your physical memory)
- if you can, use 2 different onconfig's: first when importing data and second when creating indexes and when working as warehouse. In first case, put almost all physical memory in buffers and lower SHMVIRTSIZE (i.e.
for DW use much lower buffers and much bigger SHM i.e.
- if you have IDS Enterprise Edition you can use PDQ settings to improve DW queries
- use multiple TEMP dbspaces (even if they are on the same physical device)
- if you plan to use processor affinity, look in machine notes if it supported. If it is supported, you can't use all physical processors for IDS CPU vp's - then nothing left for other things (e.g. OS ) - I suggest not to use affinity, but use 4 CPU vp's (for start)

And, post this question to www.iiug.org forum - there are plenty of experienced Informix gurus. I'm sure you'll find right solution there
Also, visit iiug conference next year
HTH