Peter's Computer News - February 23, 1997Peter's Computer News - February 23, 1997Give your SCSI a boot!
About three years ago, I ran out of room on my IDE hard disks and decided to add a 1 Gigabyte SCSI drive. What should have been a simple installation turned out to be a whole adventure. But that is another story.
Since the new drive was faster than the old IDE devices, I wanted to make it the boot device. Although I could configure the SCSI controller to be a boot device, it would not boot unless I disabled the IDE drives. A couple of calls to tech support indicated that there was no solution except to boot from the IDE.
When again I ran out of disk space, I chose to replace the aging IDE devices with newer IDE drives. Although a SCSI has better performance, an IDE device does not put as much of a strain on my budget!
On my most recent upgrade, I added a second 2GB IDE drive to my system together with the now older Gigabyte SCSI drive. The IDEs were a matched pair of Seagate ST32140A drives. What I really wanted to do was to try out NT’s RAID (stripe set) capabilities with the two drives. Problem was that the boot drive cannot be a stripe set.
Although my previous experience told me that the IDE had to be the boot device, I was still going to do some experimenting. My initial plan was to create a small boot partition on the IDE drive and have it direct the actual boot to the SCSI device. Such an arrangement is possible by using something like the OS/2 boot loader. I still have my copy of OS/2, but have found that the boot loader is the only sometimes useful part of it (I have been using NT since the SDK).
An interesting thing happened during my configuration changes. At one point, I went into CMOS and disabled one of the IDE drives by changing its setting to "none". When I booted NT, I was at first surprised to find that the "disabled" drive was still there in Disk Administrator. At first I though that I had made a mistake, but quickly realized that NT was bypassing the BIOS! NT uses its own drivers to access hardware. Since the IDE drive was partitioned and formatted, it did not matter that CMOS said there was no drive there.
Armed with this new discovery, I went back into CMOS and changed the other IDE drive setting to "none". With no IDE drives present as far as the BIOS was concerned, the boot process transferred to the SCSI BIOS. Success! I could boot from the SCSI device and still have the IDE drives present.
I should have recognized this possibility long ago, since I already knew that the CMOS settings such as LBA and drive parameters do not affect how NT sees the device. The CMOS settings are only important when using DOS or DOS based FDISK. Sometimes it takes an error to see the possibilities.
With my new configuration, I have the SCSI drive divided into two partitions. One is used for NT and the other has Windows 95 on it. No, I don’t care very much for Windows 95, but there are still some programs that just don’t NT versions yet. I do tell the vendors to hurry up and update though.
Now I seem to have the best of both worlds. I boot from the SCSI drive, but have full access to the low cost IDE drives as well. I can boot to Windows 95 or NT. Under NT, the Win 95 partition becomes drive F while the NT boot disk is C. This is possible because of the NT drive remapping capability. About the stripe set? No, I don’t use it. Turns out there was no speed advantage gained and a whole lot of liability in case of disk failure. But that is a topic for another article.