2 ms·
Why reproducible machines didn't used to be a priority (I think)
- cratermoon 4y agoFrom personal experience I can add another observation. We didn't used to have herds of identical commodity hardware machines in racks. We started with one, very expensive, computer, bespoke for the task at hand. Maybe it was financial, for which we had an IBM s/360. Maybe it was scientific, using a CDC or later, a Cray. At some point in 1993 the concept of the rack mount came along, with Compaq's ProLiant. I worked at a place in the late 90s where we had 3 machines in a rack, load balancing out website. We could get along with just 2, under normal load, so the third acted as both surge capacity and a hot spare. Whenever we needed to maintain or replace a system, we did it one at a time, during low-traffic times. We had some in-house tools that made the down time for any individual machine small enough that it was possible to get all three replaced in a day's work. Later, the concept of high-availability really got going, so we needed more than just 3, and we needed to be able to swap out more quickly. I believe I have to credit PC makers for creating the first tools for mass configuration management. The ability to set up and maintain dozens(!) of identical desktop machines across the company made life easier. Tools for replicating systems like Puppet came along in 2005 and later. It became possible to deploy racks of identical cheap machines, identically configured.