Guest Review - A Dual Drive Striped Array in a 7500 (7600, 7300)

By Mychailo Toloczko 5/18/98

Introduction

A look under the cover of a 7500 (or 7600 or 7300) suggests it was not designed with the notion that the owner might eventually wish to install a dual internal drive/PCI SCSI card array. There's no built-in way to route a ribbon cable from the PCI card area to the drive bays, and the drive bays are not designed with good cooling in mind. However, with the right drives and a little improvisation, a dual-drive/PCI SCSI array is well within reach.

Performance

I might as well start of with the good stuff. Here is the setup: 7500 w/166 MHz 604e Apple card 80 MB RAM 256 KB L2 cache one or two 2GB Atlas II drives (XP32275W) Initio Miles PCI SCSI card, BIOS 1.01, PCI slot A1 IMS TwinTurbo M4, TT402 driver, PCI slot B1 4.0 MB disk cache VM off System 8.1 Speed Doubler 8.1 Conley SoftRAID 2.0.2 Standard HFS volume Mac Bench 4.0 (MB4) disk scores for a Miles/single Atlas II system and a Miles/dual Atlas II (striped array) system are shown in Fig. 1. The dual drive DiskMix score is not shown because MB4 failed part way during the test. (After writing the test files to the array, MB4 would perform a small fraction of the test and then would abort with a "5050" error. Mike saw similar errors when he performed MB4 Disk Mix and Publishing Disk Mix tests using the Initio/Softraid dual drive combo.) All other tests with the dual drive combo went without a hitch. The single drive Disk Mix and Publishing Disk Mix scores fall a little below the MB4 scores reported by Mike (456, 382 repectively), but this is not too suprising considering his test machine utilized a 180 MHz 604e with a 1 MB L2 cache whereas my 7500 has only 166 MHz 604e with a 256 KB L2 cache. The MB4 read and write tests shown in Fig. 1 basically showed that a dual drive striped array can realize up to twice the performance of a single drive setup.

Compatibility

The Miles BIOS 1.01 is not compatible with Ram Doubler. Initio reported this to me when I was having difficulting getting my system working, and I found it to be quite true. Additionally, BIOS 1.01 also may not be completely compatible with VM, at least under System 8.1. I found that Norton Utilities 3.5.2 could not complete an examination of my dual drive array when VM was enabled, but that it successfully performed an examination with VM disabled. For the time being, I have choosen to disable VM as it is important to me that I be able to properly maintain my array. As I said, with VM disabled, Norton Utilites 3.5.2 successfully examined my array. The latest version of Disk First Aid (v8.1) also functions correctly with VM disabled (I haven't tried it with VM enabled.) I also downloaded the demo version (v1.1.4) of Disk Essentials and gave that a whirl. It incorrectly identified the array as two separate drives. As expected, an attempt to examine either drive resulted in Disk Essentials claiming the drives were extremely screwed up. I mentioned this to the folks at Total Recall, and they informed me that RAID support was the next list of tasks after completing HFS+ compatibility. With the exceptions noted above, ie MacBench Disk Mix and Norton with VM, all software seems to be working just fine with the dual drive striped array.

Installation Overview

Because of clearance problems associated with the location of the 68 pin plug on the top of the Initio PCI SCSI card, to complete this installation, you MUST grind material off the plastic swinging cover that sits above the PCI cards and processor card. I tried placing the card in each of the three PCI slots and in each instance, material needed to be removed for a clean fit. Failure to remove material could lead to significant stresses on the SCSI card and on the processor card because the plastic cover is designed to anchor the processor card. I suppose the user could simply remove the plastic cover, but its not difficult to grind the cover, and the grinding procedure has no effect on the sturdiness of the machine. I looked at the pictures of the other SCSI cards that Mike tested, and with the exception of the Atto Express PCI card, it appears grinding would be necessary if any of the other PCI SCSI cards were to be used. On the Atto card, the 68 pin plug is located on the front of the card. You will need a file or a grinding wheel. If you have the former, be prepared to spend a half hour grinding. Installation of one or more cooling fans is optional. High performance drives often emit large amounts of heat, and to keep hot running drives from overheating, good cooling is necessary. Inspection of the drive bays on a 7500 shows no way for air to circulate, suggesting less than optimal cooling conditions. However, Quantum 2GB Atlas II drives don't seem to run very hot. Rather than keep it qualitative, I decided to measure the temperature of the drives after having been run continuously for several days. Immediately after shutdown, I popped the cover and measured the temperature on the top of the drive using a thermocouple and readout. I choose this location for the thermocouples as it was easy to access and it seemed to be the warmest part of the exterior. With no additional cooling, the drives reached a peak temperature of less than 100F (I actually meastured 97F) when installed in my machine which is located in an office with temperature set to about 68F. The Atlas II spec sheet shows that these drives can operate at temperatures up to 131F. Assuming the Quantum spec sheet is quoting drive surface temperatures, then the operating temperature of the drives is well within spec. Before I got around to actually measuring the drive exterior temperatures, I choose to install a cooling fan at the front of the lower drive because, by touch, the lower drive seemed to run a bit warmer than the upper drive. The installation of this fan is outlined below.

Installation

Backup all the data on the old internal drive(s) to an archive that can be accessed after the installation is complete. I chose to backup all my data to an external HD. This is certainly the most efficient solution. An overview of the completed installation is shown in Fig. 2. The card is installed in slot A1. I choose slot A1 because I've read several times that its best to install a SCSI card in the slot closest to the processor. Several minor hurdles were encounted during the installation. The first has to do with installing the Miles card in the slot. The card I purchased has a strip of foam glued to the rear end of the "non-chip" side of the card. I'm not sure what purpose this foam strip is supposed to serve, but it hung up on the sheet metal grating at the rear of the machine when I made the first attempt at installing the card. The problem was resolved by bending over the offending grating a little bit. Next the drives were installed. 2GB Quantum Atlas II drives were used. The Initio Miles utilizes a single-ended interface, so if you are looking to duplicate the system described here, the correct Atlas II drive to purchase is the XP32275W. This is the 2GB drive with a single-ended Ultra Fast/Wide interface. The "W" denotes that its wide and single-ended. There are two sets of jumpers on these drives; the set at the front is the easiest to manipulate, and a schematic of these jumpers (borrowed from Quantum's web site) is shown in Fig. 3. The Initio Miles card supplies termination power, so on both drives, remove the jumper that enables termination power. (Save these jumpers as one will be needed later.) I'm recommending that termination power be disabled because I've been told its best to have only one source of termination power on a SCSI chain. See for instance <http://idbdnet.com/infotech/scsisolved.html>. To work properly with the Initio card, I found that the "Delay_Spin" jumper must be enabled. The drives will not mount at startup if this jumper is not enabled. If the data ribbon cable that came with the Initio card is to be used, then the upper drive will be at the end of the SCSI chain which means that the "Active Termination" jumper should be enabled on the upper drive. This jumper is next to the Delay_Spin jumper. I set the SCSI ID's on my drives to 1 (enable A0) and 2 (enable A1). The next step is to acquire a file or grinding wheel and remove a little plastic from the plastic cover that sits above the PCI slots and the processor card. It is necessary to do this because of clearance problems between the ribbon cable and the unmodified cover. I strongly recommend grinding the cover. Neglecting this step could place substantial stresses on the Miles card and also the processor card. Fig. 4 shows where I removed material from the card. To get a better feel for where material needs to be reomoved, install the ribbon cable on the Miles card and fold over the ribbon cable as shown in Fig. 2. Then swing the cover on top and look to see where it touches the ribbon cable. After grinding of the cover is complete, install the ribbon cable as shown or as you wish if you find a better way. The ribbon cable shown here is the one that comes with the Miles/Softraid package and appears to be a good quality cable. Fig. 2 shows the routing of the cable. If you are following my setup exaclty, the top drive is to be the last drive in the chain. I could have more cleanly routed the cables if the plugs were oriented differently, but that's how it goes sometimes. At first guess, it may appear that the ribbon cable would get pinched between the metal tray edge and the computer cover when its reinstalled, but inspection of the height of the tray and the cover showed there was enough room for the cable. I've examined the cable several times since installing the array, and the insulation is only slightly compressed in a few spots which indicates the cable is getting pinched slightly. In the location corressponding to where the ribbon cable will go over the tray edge, you may wish to push in the tabs on the inside of the machine cover. Timothy A. Seufert noted on the xlr8yourmac page that there is a panel, just behind the HD drives, that can be removed to route the ribbon cable. This method eliminates the chance of pinching the ribbon cable between the machine cover and the tray edge, but by removing the panel, the tray is apparently weakened somewhat which could make the case unsuitable for supporting a large monitor. Since the only SCSI device now connected to the internal stock Fast SCSI-II controller is the CDROM, the CDROM needs to be terminated. Inspection of my CDROM showed that it was a Matsushita drive. Jumpers were located at the rear, and all jumpers except for one were labeled. None of the labels showed a termination jumper. My suspicion was the unlabeled jumper was to enable termination, but I had no proof. After a long and fruitless search on the net, I still had no answer, so I punted and assumed it was for enabling termination. If your drive is the same as mine, you will need to use one of the spare term power jumpers removed from the Atlas drives as the jumpers on this CDROM are slightly larger than than average. I have no way of confirming if that was termination enable, but everything is working fine... The basic installation is now complete. If you desire to add one or more cooling fans, see below. The machine, just prior to installing the cover is shown in Fig. 5. (The cooling fan can be seen at the front of the machine.) Pop on the machine cover, and you are ready to format the drives and create a raid volume.

Optional Fan Installation

Inspection of the 7500 shows there is room to mount a small fan to the outside of the sheet metal sitting in front of the drives. To mount a fan in front of the upper drive, the fan must be no larger than 1.25 inches square and 0.25 inches deep. To mount a fan in front of the lower drive, the fan must be no larger than 1.5 inches square and 0.4 inches deep. The fan I used was 1.5 inches by 0.4 inches and it was a tight fit. The fans will need to be 12 volt DC units. As changing out a dead fan is a hassle, buy good quality fans. The fan used on the system described here is shown in Fig. 6. For a clean installation the following is required: a drill, a 3/16 inch bit, a 5/16 inch bit, four #10x3/8 inch sheet metals screws, two "scotch locks" (18-22 awg) to splice into power to the CDROM, about 5 inches of 18-22 awg dual conductor wire, and a male/female connector. Splice into the 12 volt power going to the CDROM. "+12 volts" is the yellow wire and the black wire next to it is the corresponding ground. Use a little of the dual-conductor wire to go between the spices and the female end of the connector. I attached the fan such that it would draw in fresh air and blow it over the drive rather. After attaching the fan to the sheet metal as shown, attach the male connector to the fan wires. Pop the sheet metal shield back into place (see Fig. 5) and snap on the machine cover.

Feedback

I welcome your comments. Additional suggestions and comments you wish to share with others through the xlr8yourmac site are best sent to Mike. Back to XLR8YOURMAC.COM