@peron image description: two (IBM?) technicians are working in a room full of giant tapes and one looks to be loading a tape into one of the many machines around them

@theruran @peron
<voice of="obiwan kenobi">
Those aren't tape drives. Those are hard drives.

The tape drives are along the exterior walls of the data center. The man appears to be looking for a particular volume to mount into a harddrive, while the woman appears to be in the process of (un)mounting a disk-pack from a drive, presumably the one the mainframe requested. If you zoom in on the packs and increase brightness, you can see multiple platters inside.

Each one of those is probably 7MB (IBM 2311) or 30MB (IBM 2314) or so of storage.

I count nine harddrives in the picture, which means that the mainframe can work with up to 64MB or 270MB of "directly-addressable" storage (depending on the model numbers of the drives; also, hence the term "DASD", or Direct-Address Storage Device) at any given time (assuming no additional harddrives that we can't see). Each of those drives can sustain about 5MB per second of I/O throughput to the mainframe per channel. So, if you figure two channels were used on a high-end mainframe of the time, a whopping 10MB/s data throughput could be achieved.

@vertigo @peron 10 MB/s seems pretty fast for such old technology! what era would this be? mid-60s?

@theruran @peron Yes; 1964 to 1972 or so. The IBM System/360 mainframe was released to the public in 1964.

Their I/O channels were blisteringly fast for the era. In fact, that's the key distinguishing characteristic of a mainframe: raw, unadulterated, brute I/O performance. Computationally, mainframes are quite poor in comparison to other contemporary compute devices (even today). But, if you want to push data around, nothing beats a mainframe. Nothing.

Part of the reason why IBM's I/O channels were so fast is because there was a dedicated processor for each channel on the mainframe's side of the connection, and these channels connected to controllers on the other side. Controllers were, as you might expect, also intelligent (often being smaller computers of their own). The mainframe did not often deal with low-level communications details with peripherals; instead, I/O channels were message passing in nature.

@vertigo @theruran @peron hardware addressing was terrible until you realized there was a control bit involved.

@theruran @peron Each channel moved data at about 5MB/s under ideal conditions. (Real-world performance varies, of course). On the more expensive mainframes, each channel had its own path to its own chunk of memory, which is why each channel could operate at the same time. Channels only had to fight with the CPU, not with other channels, at least when accessing its own chunk of memory.

@theruran The IBM 1441 provides computerized love for everyone!

LB: This is what a typical mainframe computer setup looked like back in the 60s. The harddrives in the picture probably provided upwards of about 270MB of storage across 9 drives. (Harddrives back then had removable media, so you could have much more than 270MB of storage in your computer shop, but you could only work with [in this case] 270MB at any given time.)

Assuming those nine harddrives were mounted on two I/O channels, you could access the drives at a data rate of around 10MB/s (5MB/s per I/O channel).

Those harddrives required people to constantly babysit them. They took up a room the size of a triage unit in a hospital. They required air conditioners about on par with a typical freezer in most fast food restaurants. And, the power bill would be in the thousands per month. In 1960s money. (This is why computer time was so precious, and was billed by the compute-hour.)

By comparison to today's technology, a home-brew hacker like me can make use of a 2GB SD card with data rates upwards of 3MB/s using a 25Mbps SPI link, and the whole thing will fit on your desk in about the space of a typical pocket calculator.

Sign in to participate in the conversation
Mastodon @ SDF

"I appreciate SDF but it's a general-purpose server and the name doesn't make it obvious that it's about art." - Eugen Rochko