RAID IIci

[mcdermd]

Those resistor packs are only about a buck a piece from places like DigiKey and Mouser.

 

[olePigeon]

Neither Digikey nor Mouser carry them. They're not a common part. 110 ohm 10 pin SIP.

 

[mcdermd]

My bad. I'm thinking the 8-pin packs.

 

[uniserver]

Wow so I was looking at your graph above,

On a percentage how much more throughput would you say your getting with the Stripe?

also, just a quick thought… would there be any heat build up concerns?

and yes that IIci case looks like it fits back together nicely, one would never know

well except the FCC, depending on how much of that metallic coating got scraped off in the mod LOL

all though, you would need some kind of special equipment to determine wether its emitting or accepting RF,

i wonder if you could hack a oscilloscope /w antenna to figure out this?

Its ok I will still keep my tinfoil hat on.

 

[olePigeon]

Now my power supply stopped working.

All the turning on/off made it not power up anymore. If I leave it alone over night, it'll work for ONE power on, then not work again until the next day.

I'm beginning to suspect that's why that Dell machine was headed for the recycler. The PSU was going. Sigh.

 

[trag]

I found SCSI termination resistor packs. Details here:

http://68kmla.org/forums/viewtopic.php?p=203746#p203746

 

[olePigeon]

Wrong resistance. Need 110 Ohm. I've searched high and low for them, and can't find any.

Also, I'm under the impression that my nubus Jackhammer card is broken anyway. I talked to Herb Johnson at Retro Technology, and he seemed to think that a 100 or 120 would probably work just fine. I tried them both, but the card still freezes up or gives me a Sad Mac with a BUS error code (presumably SCSI.)

When configured correctly with just the one resistor, it'll freeze after only a few minutes of use. So I guess the card is just broken. Keeping my eye out for another one, but they're not too common. I thought about using my toaster oven to "reflow" the board. Maybe it's a flaky solder joint somewhere.

 

[uniserver]

well you cant hurt it any more then it already is.

maybe get some flux on it and " toaster oven " that bad boy

 

[trag]

Why do you think that you need 110 ohm? Did FWB use a simple set of isolated or bussed resistors rather than a termination network? Were you able to read the part number off of an existing termination resistor pack?

If you simply took measurements on the pins of an existing termination network, I calculate that a 220/330 termination network will have a resistance of about 100 (98.5) between each of the center pins for a 10 pin network and 120 ohms between each of the center pins for an 8 pin network. Both of those are close enough to 110 ohms that you could probably get 110 ohm measurements, given the tolerances.

See page 2 of this document for the different diagrams:

http://www.bourns.com/data/global/pdfs/4300R.pdf

 

[Trash80toHP_Mini]

I've been meaning to revisit this thread, suggestion:

Remove the termination resistors and hook up an internal drive to test to see if your card is good.

If it's the other way around, attach an external device with termination.

If terminated devices on both ends (without the resistor packs on board) won't work, I'd be suspecting that the card is bad. :-/

I've got a good card with the proper terminators to test it for you.

 

[Cory5412]

As a side-note about installing a significantly more modern disk in the IIci, something like a 10k server disk from the last five to ten years, heat will be a concern if you put too many of them in (two might be too many) -- most of those disks were designed for equipment in air-cooled datacenters where the machines were designed with either more space between disks or where air flows over the disks then through the rest of the machine.

Also, a single 10k disk will probably completely saturate any NuBus SCSI controller.

 

[Trash80toHP_Mini]

. . . a single 10k disk will probably completely saturate any NuBus SCSI controller.

Yeah, I was thinking that I probably don't need to stripe two SCA drives in the StudioArray, even with NuBus 90 on the 81/110 . . . :-/ . . . but, might a Rocket doing NuBus block transfers to/from the JackHammer make setting up a second drive for RAID be worthwhile? :?:

Maybe even under the IIci's or IIcx NuBus implementation?

I'm also wondering if RAID using a pair of my 2.5" 10k U320 drives would be silly on the Fast/Narrow SCSI II bus of the Rocket's daughtercard?

 

[olePigeon]

I've been having some success with my FWB card. I went to Weird Stuff and bought some active terminators. They have boxes of them for $1 each for all different kinds. Passive, active, external, internal, etc. I also terminated the onboard SCSI at both the external and internal. After doing that, I was able to get my IIci to boot without a Sad Mac. I then even got it to boot off a 5400 RPM 68 Pin SCSI drive without freezing! Also got a centronics active terminator for my peripherals. With an active terminator, I'm going to give those other jumpers a try.

Unfortunately my last 15K RPM drive died. They keep dying on my G4. I don't know why. If I ever run into some extra money, I think I'll just bite the bullet and get a SATA card and a relatively cheap SSD for my G4. SCSI on that thing has become too expensive with my failed drives.

 

[uniserver]

With your g4 I think it would be safe to say,

Its a heat related failure, those 15k's generate some heat.

PS: do you still have a IIci for sale or trade?

 

[olePigeon]

Sorry, they were all sold.

 

[Cory5412]

Yeah, I was thinking that I probably don't need to stripe two SCA drives in the StudioArray,

This depends on a lot of things! How old are those SCA disks? If they're SCA disks from the olden days when SCA was used in things like the original SGI Octane and the O2, they may actually benefit from being run in tandem. The first generation of SCA wasn't *that* fast.

However, if you're talking about U160 or newer disks that use the SCA connector, then essentially nothing Apple or any Mac clone that's beige can gain anything from running them in RAID from the onboard controller. If you have a good PCI controller, then it may be worth your time.

For what it's worth, in fact, I would be completely unsurprised that a fast 10/15k 2.5-inch U160 or U320 is faster than NuBus itself. But we're talking about a disparity of ten to fifteen years between "system" and "disks."

If I ever run into some extra money, I think I'll just bite the bullet and get a SATA card and a relatively cheap SSD for my G4.

This is probably what I'd do if I had a Power Macintosh G4 on hand. Even a fresh 5400/5900 or 7200rpm spinning disk will be a large improvement over anything IDE, and if your scsi disks are dying, it may be faster than they are, too.

those 15k's generate some heat.

And of all the Macs ever, I'd argue that most Power Mac G4s are better at having some air to begin with, and pushing it through the enclosure. This is especially true in comparison to some of the more compact 68k Macs, such as the IIci.

If you wanted to be safe and have a number of fast old hot SCSI disks, you may look into something like one of the old Sun STOR-EDGE shelves or something like the Sun 711 disk pack, which had reasonably good cooling for what it presumed at the time was up to twelve half or full-height disks. You may even consider mounting them in an old AT PC case, or grabbing something like this if adequate cooling for a lot of SCSI disks is a priority.

One Mac I'd be interested in seeing the exact thermal performance of, or maybe just having one or two of so I could test this theory out, is the Quadra 950/AWGS95. The AWGS9150 would also be fun, and the combination of the 9150, system 8.1, and ASIP5 or ASIP6 could make a particularly good old server, especially given that the 9150 and 950 seem like they have a reasonable amount of air and a reasonable path for air to flow over the disks and out the back of the machine. Although I haven't looked in particular.

As a sidenote, one entertaining thing I've long wanted to do is grab an old disk shelf from a server (Something like any older Dell PowerVault enclosure, the 630f would be a good example) and rip most of its guts out, leaving just the structure for disk mounting, power and some fans, and adapting it to the task of holding modern disks.

Good luck on your SCSI adventures!

 

[Trash80toHP_Mini]

This depends on a lot of things! How old are those SCA disks? If they're SCA disks from the olden days when SCA was used in things like the original SGI Octane and the O2, they may actually benefit from being run in tandem. The first generation of SCA wasn't *that* fast.

I'm talking about the 2.5" 10K U320 IBM/Seagate Savvio Drives we were ironing out termination issues with a bit ago. DOM: 06SEP2005

Specifically, I'm talking about:

1) Running them fast/wide off the JackHammer swappin' spit via NuBus Block transfer mode in the Radius 81/110 with a VideoVision Studio Card . . .

. . . I think that's how the VideoVision StudioArray works . . . [:I] ]'>

2) Running one (or two?) fast/narrow on a Rocket's '040 PDS SCSI II DaughterCard at the 33MHz system bus rate.

 

[Bunsen]

at the 33MHz system bus rate.

Interesting... that's an '040 PDS (more or less). I doubt it would make much difference really, but interesting nonetheless.

 

[Trash80toHP_Mini]

I would think it ought to be a fairly significant difference in throughput, maybe even a near equalizer against the JackHammer.

Just like piddling PDS VidCard/MoBo Video is faster than accelerated NuBus VidCard behemoths at smaller resolutions, the extra bandwidth of the '040 PDS on the Rocket should prove interesting for benchmarking against the NuBus JackHammer. Does the digital water flow more quickly at 16bit through the NuBus funnel at 10MHz or the 8bit funnel at the higher 33MHz flow rate of the '040 PDS.

16Gal at 10MHZ vs. 8Gal @ 33MHz

160 GPM vs. 264 GPM

Only the sha . . . erm . . . benchmnark know . . .

 

[Cory5412]

A single u320 disk should work well for this and completely saturate all available throughput regardless of how you hook it up. I'll have to look at these disks for the old DEC 3000 I've got hanging around, too. A single one shouldn't be toooooo bad.