Announced in about 1968, the IBM
2301 Drum Storage Unit was part of IBM's System/360
It replaced the earlier 7320 Drum Storage Unit.
With a price tag of about $80,000US and a total capacity of about 4MB
, these were not
the devices that data
belonging to mere mortal users
got stored on.
Instead, 2301s were used as paging devices
on the IBM mainframes
of the era.
With an average access time of 8.6 milliseconds and a data transfer rate of 1.2 megabytes per second, these were the speed demons of their day.
A 2301's data was stored on a drum or cylinder about two feet high and about a foot in diameter.
The data was stored in parallel tracks that ran around the outer surface of the drum.
Each physical track had a dedicated read-write head, a feature which totally eliminated seek latency because the device could instantly switch between tracks electronically (seek latency is the time spent while the read-write head on a conventional disk drive moves to the required track).
To improve the data transfer rate, four physical tracks were grouped together to form a single logical track.
Each logical track could hold about 20,000 bytes (my recollection is that the formatted or usable capacity was about 19,000 bytes).
With a total of 204 logical tracks, the 2301 had a capacity of 4.09MB.
The drum had a rotation time of 17.5 milliseconds (the time required to rotate once) and an average rotational latency of 8.6 milliseconds (the average amount of time that it took for the desired data to come under the track's
The maximum data transfer rate of the device was 1.2MB per second.
Physically, an IBM 2301 was a sight to behold.
The unit stood about six feet and was about three by three feet square.
The top third of the unit had glass windows on two sides (front and back)
which allowed one to marvel at the silver coloured drum which
was oriented vertically and surrounded by the read-write heads
(arranged in columns around the drum and each with banks of
coloured wires coming out of them).
The bottom two thirds of the unit contained the electronics (hidden from view behind metal panels).
A maximum of two 2301s could be attached to an IBM 2820 Storage Control Unit.
A maximum of two 2820s could then be attached to a
channel on the mainframe.
Each 2820 was about six feet high and had a footprint of about three by six feet (when placing them into a machine room, enough space had to be left around each 2820 to be able to open the side panel doors for maintenance purposes).
There was a significant limitation that needed to be taken into account when attaching 2820's to a mainframe - the 2820 required that the maximum cable length from the mainframe's channel port to the 2820 be about 20 feet (if two 2820s were connected to a single channel then the length of the channel cable between the mainframe and the first 2820 plus the length of the channel cable from the first 2820 to the second couldn't exceed about 20 feet).
In one mainframe computer room that I worked in, they had thirteen 2301s connected via seven 2820s to four channels on a single mainframe.
Take this cable length limit, the footprints of the 2301s and the 2820s, and the space required around them for maintenance purposes and you've got a really crowded herd of boxes by the mainframe!
The IBM 2301 was the fastest or one of the fastest computer storage devices in the world when it was announced in about 1968 (does anyone know the exact date?).
Although other higher speed storage products, like the IBM 2305 fixed disk storage unit, would be announced later,
the 2301 would continue to be a viable choice
of high speed storage for almost fifteen years.
The computer room mentioned above had thirteen 2301s and seven 2820s providing paging space in about 1982;
a 2305 was faster but the paging capacity provided by thirteen used 2301s with control units cost a lot less than the same capacity provided by used 2305s and there weren't any other alternatives available (solid-state paging devices (big boxes filled with DRAM) started to appear in about 1982).
IBM also sold a 2302 and a 2303 drum storage unit.
I don't have any details on these other than that they were similar to a 2301 but had different capacities and performance characteristics.
The test of time
Finally, let's see how the 2301 measures up against today's hard drive technology.
The IBM 2301 had rather limited capacity by today's standards at 4.09MB (less than four floppy disks
The data transfer rate wasn't very fast either at a mere 1.2MB per second (it was about 50% of the peek transfer rate supported by IBM mainframe channels of the day).
However, the average access time of 8.6 milliseconds has stood the test of time rather well:
- IBM's current top-end consumer-grade drive, the 120GXP has an average seek time of 8.5 milliseconds and an average access time of about 8.6 milliseconds.
IBM's top-of-the-line 146Z10 Ultra320 SCSI drive which is intended for server applications has an average seek time of 4.7 milliseconds and an average access time of 4.8 milliseconds.
This the test of time section was written on 2002/10/02.
According to IBM's web site on that date, the 120GXP and 146Z10 were IBM's top-end drives in their respective markets.
A last set of numbers (to ensure that the basis for the above comparison is recorded):
- the IBM 120GXP has an average seek time of 8.5 milliseconds and, with a rotational speed of 7200 RPM, a rotational latency of 0.14 milliseconds.
- the IBM 146Z10 has an average seek time of 4.7 milliseconds and, with a rotational speed of 10,000 RPM, a rotational latency of 0.10 milliseconds.
Sources: mostly personal recollection with a few details taken from various
web sites including:
- "Notable Events in the Storage and I.T. Industry (A Historical Summary)"
web page located at
http://www.wwpi.com/lead_stories/Notable_Events.htm (last accessed 2002/10/02)
- untitled web page located at http://www.dorje.com:8080/netstuff/folklore/old.machine4 (last accessed 2002/10/02)
- untitled web page located at http://www.cs.niu.edu/~mstack/assist/assist.tape/ASASSIGN (last accessed 2002/10/02)
- IBM's web site at http://www.ibm.com/ for specs on the 120GXP and 146Z10 (last accessed 2002/10/02)
- "General Information Manual / Introduction to IBM Data Processing Systems"; Copyright © 1960, 1964 International Business Machines Corporation (this is a manual in my collection which contains a photograph of a 7320 Drum Storage Unit)
There's a rather grainy photo of a 2301 at http://www.columbia.edu/acis/history/drum.html (does anyone have a better one?)