When I started on the server team at my company in the mid-90’s the client server model was in full swing. We were using Netware version 3.x. Our network environment was token-ring on a FDDI fiber backbone (that was considered very advanced at the time). The term distributed computing on the intel platform was born from the servers having to be distributed in the office. Latency caused servers to be located near the departments.
The Finance server was typically located in the Finance department. The servers were usually located in a locked closet with the network switch equipment. Whenever a server went down we need to know where the server was located (3rd floor south wing), we needed to make sure we had the key to the closet, and if it was a campus setup you had to walk over to a different building.
We have come a long ways since distributed computing was really “distributed”. The first step was going to ethernet for networking which allowed us to collapse the server farm into a single datacenter at a campus location. For many of us this moved our Wintel servers into the same raised-floor space as the mainframe. This provided a new level of centralization, but it was still necessary to host the server farm at individual campus location due to networking limitations.
Then came the 2000‘s, through mergers and acquisitions our organization acquired three other datacenters. It also provided WAN speeds fast enough so that we could centralize all our servers to a single campus location. In our case the servers were in Columbia, South Carolina. This was a big change not only from a physical standpoint, but also from a operational standpoint. When the servers were hosted at each campus location the IT staff at each site had their own set of rules. The servers in Boston were built different then the servers in Chattanooga, Portland, and Columbia. The IT staff at each campus office location were very protective of their individual server farms and didn’t always have access to each others resources (this included the Half-Life server we had running in the datacenter for bandwidth testing).
This all changed when the Wintel servers were collapsed to a single datacenter. Builds became standardized, support tools became unified, and the operations staff for each site were all given broad authority to support the servers.
Sidebar: Going to ethernet!
Anyone who has been in IT long enough still remembers the conversion from token-ring to ethernet. Principally because we had to visit all the desktops in the organization (we had 2200 IBM desktops) and replace the token-ring cards with ethernet cards. It took us months to get through all the machines before we could even make the transition to ethernet!