Old school backup systems: what are they?

Question

So being a sysadmin for several years, I learned that "Jesus saves, but God makes tape backups," several times over.

Watching Season 2 of Halt and Catch Fire, there is some data loss.

Season two, episode three, Gordon unleashes Sonaris on the Mutiny network, wiping not only the corporate drives, but some customers' as well

Being that the show takes place in the mid 1980s, what were their options for (potentially centralized) backups of production data? There's some allusion to re-installing the backup drives; I assumed these were dumb, block-level clones taken when there was some extra slack in the network, not some good managed backup framework with fancy things like deduplication or application snapshots.

Answer 1

TL;DR: Backup (of local data) at the time the show is set was mostly done to floppy disks, and even when using a network, all was at the mercy of the user to remember doing it (some companies tied backup into system boot, but then there are zillions of stories about users disabling it, 'cause waiting sucks).

Backup of centralized data was done independant and with whatever the mini or mainframe provided (usually tapes).

---

Being that the show takes place in the mid 1980s, what were their options for (potentially centralized) backups of production data?

Quite similar to today, well except large disk arrays that is. Tapes where the choice since the mid 70s (when they slowly came out of fashion as main storage). Since then they ofc improved quite a lot. Where originally 6250 bpi tapes where standard, the mid 70s brought the introduction of various cassette based tape systems. While there was quite some competition, the QIC standard did gather the largest market share in the non proprietary market (What a bunch of words to say that IBM didn't care and developed their own standards (LTO)).

During the early 80s, which is the seting for the show, four and nine track QIC drives where most common, used with 300, 450 and 600 ft tapes. Cassettes where named according to the tape length, so a DC300 held a 300 ft tape. Four track QIC drives stored about 20 MiB on a 300 ft tape, while the nine track variant doubled this to 40 MiB (plus better error correction).

At said timeframe these drives where mostly used in (UNIX) minicomputers and in top end workstations (Apollo for example delivered most machines with a QIC drive). For simple PCs, such drives where available but rather rare. Keep in mind that during the first half of the 80s, floppy-only machines were still common. It wasn't until 1983 that IBM introduced the XT with a standard option for a hard disk.

I guess that the part of the show where next to every machine has a HD is less historically correct, but necessary to make it belivable to today's viewers :)

Besides QIC, there where other standards which did get some audience during early PC times. For example DCC, working on compact cassete like drives, but using a digital format with special tapes. They could store similar amounts per cassette, but has slower access times.

So yes, single machines could have been equipped with tape drive (QIC or other), but it's rather unlikely that it got rolled out all across a company. In context of the show, some introduction of a company invented new device would make sense... much like the Iomega did with the Ditto drives in the 90s.

The standard medium for backup were floppy disks. From today's view this may sound impossible, when a single letter can have several megabytes, and a floppy just holds 360 KiB, but back then programs didn't use bloated markup. Even with a lot of formatting a word processor file stayed under twice the amount of visible characters. A rule of thumb for letters were about 2-3 KiB per page. So one floppy could well store a hundred pages without compression or alike. Quite workable for a weekly differential storage for a single PC.

Bottom line, while there have been solutions available even before the IBM PC got introduced, the real need for tape drives for single users didn't manifest before the mid 90s, when disk sizes and more importantly data formats increased exponentially in size.

There's some allusion to re-installing the backup drives; I assumed these were dumb, block-level clones taken when there was some extra slack in the network, not some good managed backup framework with fancy things like deduplication or application snapshots.

Why not?

Well, I have to admit, I neither watched the show closely nor do I have an idea what kind of network / mini / mainframe setup is featured. But since the premise is about a PC company, it's safe to assume that there is a lot of workstation orientated application and data storage, so decentralized data (data stored on each workstation) needs to be backed up.

Considering a network with most PCs connected (rather unusual for the early 1980s) it's quite possible that there was some IBM mainframe with extensive tape storage used to backup workstations. Usually the backup would be made in two stages: first the single PCs would be backed up onto mainframe disks with separate areas for each user, and in a second step all these user backups would be exported to tape.

block-level clones

Most likely not.

The backups in step one would NOT be made up from simple block copies, but rather on file level and differential, meaning that only files changed since last backup will be copied. DOS supports this strategy since 1.0 via the Archive bit, inherited from CP/M 2.2.

taken when there was some extra slack in the network

Not in the early 80s and on PCs.

MS-DOS was a single user single task operating system. There was no room for a backgound process nor the necessary interfaces to do a backup without interfering with the user. DOS wasn't reentrant and the archive bit mechanics did not work well in a multi-tasking environment.

Backup was strictly a process done at will and never really in the background. Considering the meager amounts of data to back up, even over a slow network all transfer could be done during a day in a few minutes. So no need for additional complexity.

Of course one could assume they were using some kind of multitasking environment running every PC, but that's extremely far fetched, as Desq (later named DESQview) wasn't introduced until 1984 and at first was 'just' a switcher on top of DOS, thus not able to run an independent background process. The 386 changed this, but that again should be way beyond the time frame the show is set.

Similar for Novell's Netware (to consider the network side). Until Netware 286 V2.0 (?1986 or 87?) it was a strict add on to DOS, hooking some functions to redirect access to a specified 'drive' over the network. All activity was client initiated. Again, no background processes at all, thus none waiting for 'network slack' to do some background.

And yes, there are zillions of other posibilities and combinations, I tried to stay with the most common/generic.

Answer 2

Adding to Raffzhan's answer, many obscure, short lived and slightly crazy data storage systems were created during the 1980s due to the high cost of mass storage.

For example, several systems were created to save data to video tape (usually VHS format), because tapes and VCRs were widely available and relatively cheap. They worked somewhat like audio cassette data storage that was popular on 8 bit computers.

People tried printing data on paper and scanning it back in with special low cost scanner hardware too.

Answer 3

Back in the mid-70s (obviously earlier than the HCF setting), we just used disks to backup disks on the Apple IIs the teachers used at (secondary) school - you did say old school, didn't you?

Since our nerd cadre was in good with the maths teachers, we were tasked to write an automated tool to actually do this for the staff. You would basically insert the backup tool to run then, each original and backup disk would be placed into the two drives. It would simply erse the current backup then copy all the files one by one.

From memory, these backup disks were then labelled and kept in an inflammable cardboard box near the staff heater :-)

Answer 4

QIC drives for the 1980's PC were often incredibly slow, writing at speeds of approximately 250-500kb/sec. There was an adapter from a company called Colorado to plug them directly into a floppy cable, limiting its speed to that of the floppy controller.

One of the main "features" of QIC was its notorious unreliability. Writing out a QIC backup was already painfully slow, but there was no automatic read-after-write quality control built into QIC drives, so it would be necessary to followup a backup with a full read of what was just written to compare it against the original media, verifying it was written correctly.

This then effectively doubled the time needed to complete a backup job, and since the concept of "snapshots" did not exist yet, the source media being backed up had to effectively sit completely idle and unused through the entire write and verify process.

The QIC tape system was not sealed to keep out dust, so over time the head and the tape would also accumulate dust and debris, that would interfere with read/write quality.

If a followup verify-after-write operation and regularly scheduled head cleaning was not done, a QIC drive might very well just be writing out unreadable garbage that would not be discovered until it is too late.

,

By comparison, modern Linear Tape Open (LTO) does not require the full manual media verify after a write job completes, as the drive is designed to automatically read data as it is written, to immediately and continuously check that it is working correctly. The tape drive can then perform corrective measures by itself to automatically rewrite the media to recover from write errors, and if it continues to have tape verification problems, it can automatically request cleaning of the tape drive.

...though still as with QIC, LTO tape and the drive interior are not also sealed and gasketed, to protect it from dust buildup from the cooling airflow.