SkyLab Lore & eBay piccies . . .

[Trash80toHP_Mini]

. . . of an auction long ago in a place far, far away.

I'll have to search out the posts here and on 'fritter about this auction. IIRC this was in my last months in the Big Apple and I was likely more than a bit psychotic (undiagnosed manic episode) at the time. The search may prove interesting . . . or not. =8-\

Radius SkyLab















LEM article: Rocket Science 101

 

[tt]

Where are all the NuBus slots mentioned in the article? I have a Rocket I bought a while ago to play with, not sure what I would use it for. Suggestions?

 

[Trash80toHP_Mini]

My guess is that those two PDS lookin' EuroDIN connectors on the underside were for a pair of, or one very large, NuBus Risers for the Rockets. I've never actually counted pins, but the rectanglular cross section says PDS pincount to me.

With one stack of fourteen Rockets, that'd mean the PSU(s)(?) and Drive Bays would stack front-n-back on the other "side" of the Rockets above the MoBo.

Under RocketShare with the DSP DaughterCard, or the more affordable STORM Card, in DSP support, coupled with a PAS16, SAII Card or equivalent, I think an early nineties IIfx MPEGDEC stereo output CoPro rig would BD'Bombe. [}] ]'> PixelPlaytime in PhotoShop on another Rocket using the DSPs on the NuBus Card for a 21" Display with your playlist/panel in a window or on a second monitor along with your menus would be killer. you'd still have slots for a SCSI II Card and NIC left over in that config.

A custom multihreaded parallel co-processing App running multiple Rockets under A/UX? SETI work units? All new driver setup for 'NIX though.

Dunno, I'm a knuckle dragger, programming is beyond my purview, but that'd be the place for hackin' a Rocket, it never happened other than Radius' RIP code back in the day as I recall. Having that code re-surface would be better'n having a complete SkyLab turn up, especially the source code! [] ]'>

 

[CelGen]

Are these the only photographs known to exist?

 

[TylerEss]

Having that code re-surface would be better'n having a complete SkyLab turn up, especially the source code! [] ]'>

I think it might not be quite as amazing as it seems, Trash-80. From what little I know of AppleEvents, one could cook up a parallel-processing application on System 7 using all supported interfaces. Then you'd just need a big old stack of Macs connected together with AppleTalk, and you could have at it.

Such a "MAC CLUSTER" in a box full of Rockets would have obvious advantages because of the speed and latency of the AppleTalk connection. But if you were doing something sufficiently compute heavy and embarrassingly parallel, even LocalTalk might be OK.

As an example, maybe a parallel implementation of a really naive prime number test? Evenly divide up the integers between 1 and SQRT(N) between all the nodes in the cluster. Each node attempts to divide N by all the integers in its allotment. Virtually zero network traffic necessary, and it'd scale almost linearly with the number of nodes in the cluster. It would also be useless except as a proof of concept.

Heck, a skilled programmer who wanted to do it could write a Photoshop plugin to use a farm of Rockets for things like gaussian blur. You'd have to send the hunks of image data back and forth across the AppleTalk, so it wouldn't be very efficient, but it'd be a really impressive hack!

 

[beachycove]

I seem to recall that Adobe Premiere had this capability built in way back around the Quadra era.

 

[Trash80toHP_Mini]

Are these the only photographs known to exist?

I don't really know, but I haven't searched for any, more may have turned up since the 1994+- time frame.

HRMMM? < . . . fires up google . . . >

p. 1 hits:

http://www.linkedin.com/pub/jim-wolcott/4/ba4/a32 may be well worth contacting!

http://www.mail-archive.com/rocketeer@mail.maclaunch.com/msg00200.html

http://www.cbronline.com/news/radius_skylab_multimedia_server_to_pack_12_rockets

http://www.applefritter.com/cgi-bin/YaBB/YaBB.pl?board=hacks;action=display;num=1075668076;start=15

Cinemo's 68kMLA thread/post linked at the end of my thread seems to be gone, was it lost in the great crash?

pp. 2 & 3 hits:

http://www.kennedybrandt.com/supermac_insider/history/index.html

http://mail-index.netbsd.org/port-mac68k/2004/01/19/0003.html

It looks like someone else came up with the Rocket/MPEGDEC angle, I don't ever remember seeing this article.

http://new.lowendmac.com/sable/05/1213.html

Google Image Search turned up bopkes, BTW. :-/

I posted about this auction 4 mos. before I had to shut down that lifetime and flee to NC to recover at my brother's place, definitely psychotic in retrospect. Thank God for supportive family. I've had to do the Phoenix gig three times already and hopefully this will be the last of my series of lifetimes. Take good care of yourselves gang, a mind is a terrible thing to lose.

 

[KennyPowers]

Such a "MAC CLUSTER" in a box full of Rockets would have obvious advantages because of the speed and latency of the AppleTalk connection. But if you were doing something sufficiently compute heavy and embarrassingly parallel, even LocalTalk might be OK.

Resurrecting this thread because I recently attempted something similar to this, but was unsuccessful. I have a IIci with a Daystar 040, an original Rocket, and a Stage II Rocket, running System 7.1. I found this parallel fractal-drawing demo of the MacMPI parallel clustering tech developed at UCLA(?) way back when:

Other Software

It says it uses AppleTalk to communicate between each machine in the cluster, and I was like "I think the Rockets also use AppleTalk to communicate...maybe I could run this demo in parallel using the Rockets?" They provide instructions for setting everything up on pre-OS8 68k systems:

nappleseed_report.html

A single instance of the fractal program runs fine on the host machine and each Rocket in isolation (even all 3 simultaneously). However, when I follow all of their instructions (enable program linking, give every user permission to use program linking, create the proper nodelist files, etc) the "master node" always says it can't find the slave node (I tried it with just a single master and single slave before trying all 3). The slave nodes are specified by the computer's (or in this case, the Rocket's) AppleTalk name, and I've verified that they match each "computer's" name in the Sharing Setup CP. I've also tried every possible master/slave combination between the IIci and both Rockets. If I enable filesharing on the IIci, the Rockets can see it in the chooser. Am I just not understanding something about how the Rockets use AppleTalk? Maybe this particular fractal-drawing program only works with AppleTalk over ethernet? Does someone else out there with a Rocket want to give this a try?