-----Original Message-----
From: Jon Elson <el...@pico-systems.com>
Sent: Tuesday, October 01, 2024 4:55 PM
To: Tom Gardner via cctalk <cctalk@classiccmp.org>
Subject: [cctalk] Re: Might be antique computer parts
>
> On 10/1/24 18:29, Tom Gardner via cctalk wrote:
>>
>> I wouldn't call the 2314 low tech - it was the highest areal density at the
>> time, a breakthru with ferrite heads and very low cost to manufacture. Note
>> I said cost, its profit margin was enormous, in part by putting as much
>> expensive electronics as possible in the control unit. ??
>> Actually the 2314 did not ship with the first 360's in 1965; it was
>> announced in April 1965 about 1 year after the 360 announcement and AFAICT
>> from Bitsavers document dates it didn't ship until late 1966, which FWIW, at
>> the Computer History Museum, 1966 is also the date for first shipment of the
>> 2414 and its ferrite heads. BTW the hydraulic actuator design goes back to
>> the 1311 - more or less the same actuator in the 1311, 2311 and 2314.
>
> Well, yes, and in the days of SLT logic, everything was expensive. So,
> putting as much of the functions in the control unit rather than the drive
> was good. But, one thing that this mindset caused was that they could not
> have one drive seeking while another drive was transferring. The entire
> operation, cylinder seek, rotational seek and data transfer was all one
> atomic operation. That really killed the throughput of the whole disk
> system. The reason was that the IBM developers came from systems like 7070
> and 7090 where all permanent storage was on tape, and they didn't quite "get"
> how central disks were going to be to the 360 systems. They had the CKD
> scheme, where you could search several cylinders for a match of some
> arbitrary field in the DATA portion of a sector, but this resulted in massive
> slowdown of the system, as it tied up not only the drive, but the controller
> and the channel as well! Thus the need for the database system, which would
> make selecting the desired record much faster.
>
> I didn't mean that the 2314 DISK was low tech, just that the drive, itself,
> was quite spartan.
>
> Jon
For the earlier 1311, lack of overlap made perfect sense. After all, the 1620
has no interrupts, no parallelism of any kind: every I/O operation stalls the
CPU until the operation is finished. (That and the BB instruction are among
the reasons why Dijkstra rejected the 1620.)
Speaking of high profit margins: on the 1620, there was an extra cost option
called "direct seek". I don't know if involved a jumper cut or some actual
circuitry (an adder, most likely). We didn't have that, and the result is that
a seek from cylinder x to cylinder y was done by a full retract to cylinder 0,
followed by a seek out to y. It was amusing to watch the shaking resulting
from a simple "incrementing seek test" -- seek to cylinder i for i = 0 to 99.
Those last few seeks would take the better part of a second.
paul
