RAID 1.5 With IDE: Added Value Or Eyewash?

Conclusion: Pros And Cons For RAID 1.5 Balance Out

Unfortunately, the RAID controllers on DFI's motherboards won't let you reach the holy grail of optimal performance and perfect data security with only two hard drives. RAID 1.5 can't dispel the disadvantages of having redundant data storage with RAID 1: on the one hand, there's the small usable net capacity; on the other, there's no sign of any real gain in performance.

In theory, RAID 1.5 is a better performer than RAID 1 because it can read from two hard drives, just because you access the data using the same zipper-type method as with RAID 0 (striping). While we could not detect higher transfer rates with RAID 1.5, CPU utilization was certainly lower and it outperformed in the fileserver and webserver benchmarks that required good I/O performance. The main reason for this good showing is that RAID 1.5 requires a lower CPU utilization. Additional positive side effects of RAID 1.5 are somewhat shorter access times.

A serious disadvantage compared to RAID 1, however, is the fact that (as with RAID 0) you can't read any drive separately with a standard controller. The data stocks can be restored all the same because of the parity information on the hard drive. You'll need to have HighPoint's HPT372N for the restoration, though, since no other controller can handle RAID 1.5 - and if it can, then only in the nested RAID setup we mentioned above (combination of two RAID modes).

Finally, the benefits for home use are interesting, but are they strong enough to justify creating a RAID array based on HighPoint/ DFI's RAID 1.5? Overall performance is only slightly improved, while data security actually increases by redundant storage onto two hard drives. The main reason for having "secure" RAID is, after all, optimized failure safety and/ or data security - and this is where RAID 1 is equally good and supported by a larger variety of controllers. Power users are best served, as before, with RAID 0, while genuine security coupled with higher performance is still associated with higher costs and, consequently, with RAID 10 or RAID 5.

RAID 1.5 might still be an attractive option in the lower server segment (webservers), though, as the test results prove that the test system with RAID 1.5 is actually better than a simple RAID 0 in terms of I/O performance. We'll have a comprehensive test in late summer that will show to what extent the IDE RAID controller will be able to take on the SCSI-dominated segment.

  • lifeisstudy
    So RAID 1.5 won't work because the heads of the disk still need to move to the next data block to read... If you skip blocks, you still have to move past them.

    What about RAID 1.5 with Solid State disks. You have nothing to move, yet you can double your read speed with mirroring. Sounds perfect for server caches.