im looking for something that will do raid 5 with at least 4 IDE connections

blazing fast omgwtfbbq speed isnt a priority because it is just going to be a basic file server

server hardware: dual PIII 850 MHz with 768 MB of RAM

the raid wont run the os and it will be on a UPS

ill be using 4 7200.10 200 GB seagates

any suggestions? im looking at this one: http://www.newegg.com/Product/Product.aspx?Item=N82E16816102036

how significant is hardware based vs software based with the server specs i have?

5/8/2007 2:24:54 PM

The LSI MegaRaid controllers were alright. You might want to search Tom's Hardware, etc. for a showdown between it and the Promise controller that you linked.

Example: http://cgi.ebay.com/LSI-MegaRAID-i4-IDE-RAID-Controller-Series-511-4-Ch_W0QQitemZ180113469195QQcmdZViewItem

5/8/2007 5:38:38 PM

Any low end (read: 4-port) IDE RAID controller is going to be driver-assisted software RAID, so you might as well use software RAID if it is available in your configuration, unless software RAID means dynamic disks to you, in which case you should probably subcontract work like this

Assuming a Linux fileserver, I'd set up mirrored boot partitions with a failover grub config, use LVM2 on a RAID-5 set to create your root and data partitions, and party on. You can even do this all to a currently running OS image by adding 3 drives, creating the RAID-5 set as degraded, copy the data, pivotfs, and resync the array to the original drive.

For best results use 4 seperate chipset-integrated IDE channels to avoid PCI bus bottlenecks. If you use 4 drives as master/slave of 2 channels, or off of a 32 bit 33MHz PCI bus your array is going to be susceptible to multiple drive failure during the extremely long rebuild times which are consequent of such bandwidth constrained configurations -- the array will likely suffer a second failure during rebuild due to I/O timeout if it ever encounters a first. That is, if it doesn't suffer 2-drive failure in the master/slave configuration to begin with.

[Edited on May 9, 2007 at 6:46 PM. Reason : .]

5/9/2007 6:39:49 PM