Drum memory is a type of computer memory that used a drum-based design to hold the memory’s platters and read-write units; it primarily was used in the 1950s. To use a drum memory, operators had to load punch cards into each head and the programs would only work after all the punch cards were loaded. Later in the drum memory’s life, programmers discovered how to optimize the drum’s timing to make it easier to load memory. Unlike modern hard drive units, the drum contained read-write units for each platter, so there were no parts to move the platters.
The drum memory unit largely was used during the 1950s, though it also saw some use in the 1960s. While this memory is largely inferior by modern standards, it presented several advantages over earlier memory types. It was easier to load punch cards, the memory functioned several times faster and the unit itself was smaller than earlier memory types. This memory became obsolete because of semiconductor memory, which has better power and storage.
There were several slots in a drum memory unit and each slot represented a different platter for the memory. To load a program or document, operators had to place punch cards in these slots, one after another. If the punch cards were not loaded in the correct slots, then this could cause a problem that would keep the program or document from correctly loading. The punch cards are magnetic, and this memory was built from ferromagnetic materials to help record memory.
At first, operators would load a punch card into the drum memory and wait for the computer to recognize the memory before placing another card. Later on, operators were able to make programs to help optimize loading time. This program would estimate how long it would take for the memory to load the card and, when it was nearly loaded, someone was near the memory ready to load the next card. So many punch cards were needed for basic applications that this greatly saved time.
With modern memory units, such as semiconductor memory, there is only one read-write head and the platters in the memory are programmed to move around so the head can work with all the platters at once. In drum memory, there was a read-write head per platter, so there was no need for the platters to move around. This may seem more advanced, but it increased the cost of the memory, both in terms of memory and expense.