David Empson wrote:
Michael J. Mahon <mjmahon@aol.com> wrote:David Empson wrote:One similar product I can think of is the Nestar networking system we had in high school (I was there from 1981 to 1985). It had a parallel network interface between the computers, with a dedicated Apple II+ file server with a 20 MB hard disk (which was about two feet wide, one foot high and four feet deep). Around 1984 I reverse engineered the Nestar card firmware and patched it to add support for ProDOS. The Apple II card only supported Pascal and DOS 3.3, and there was probably some CP/M support (which I never investigated). The server also supported SOS, so presumably there was an Apple /// version of the card. I doubt the Apple II card would have worked in an Apple /// because it had I/O mapped into the $CFxx page. The Nestar card was purely a network interface (and could boot the Apple II). The server was running a heavily patched version of Pascal, with a special controller for the hard drive (from which it was able to boot). There was also a dual 8 inch floppy drive, which was also gigantic. The server hard drive was able to be partitioned into multiple logical volumes, which were named (in a hierarchical volume file system, I think) and tagged with the type of file system they contained (Pascal, DOS 3.3, CP/M, SOS or various special types). Partitions could be dynamially created and deleted, and they had a reasonably extensive user authentication and permission system. At its peak, our network included about 17 Apple IIe workstations on the Nestar network.Just out of curiosity, what data bandwidth did the network support?It was an 8-bit parallel bus, with handshaking. Somewhat like a parallel printer interface, with a data strobe and an acknowledge. I don't recall details about the protocol and I'm not sure if I ever analysed its performance in close detail.
Interesting that they chose a parallel network connection. That reminds me of IEEE-488.
Each sector or block was transferred over the network as a single burst, and the average performance of the network was pretty bad with a lot of computers tring to use it at once, probably due to the contention phase and polling time. Some specialised applications were able to hog the network for short periods in order to transfer a larger amount of data. One example was the network login screen, which loaded a fairly large amount of data quickly. The network card clicked the speaker every few bytes transferred, resulting in a characteristic tone as it transferred large bursts of data. My recollection is that it was comparable to a 5.25" floppy, but without the overhead of seek times, as long as the network wasn't congested. That would make it about 20 KB per second on average, with the peak transfer rate perhaps approaching 50 KB per second.
Sounds about right for a software handshake per byte. Going semi- synchronous (with sunc every few bytes) would have sped things up by about 2x. Sounds like a collision-avoidance algorithm would have helped a lot with average load, too. NadaNet averages better than 10KB/s using all software over a serial bus, and seems to hold up to well over 80% utilization. -michael NadaNet file server for Apple II computers! Home page: http://members.aol.com/MJMahon/ "The wastebasket is our most important design tool--and it's seriously underused."