First there was the WAN, then the LAN- and we integrated all the computers into networks. I loved this stuff in the 1980s and early 1990s. I could at one time administer LanMan, Banyan, Novell, 3Com, and several others. I was deep into the protocols for Ethernet, Token Ring, Token Bus, FDDI, XNS, NetBIOS-NetBEUI, SNA, TCP/IP, spanning trees and more. I optimized some WAN topologies, frame relay, did tradeoff studies and provisioning, and implemented fault tollerance and backups. The DSU/CSU, firewall, brouter, hub, switch and router were my friends.
Enterprise Data Model:
At one time the "big idea" in integration was that all the enterprise applications could share a big, central database with a single, large data model. This idea eventually eventualy became the ERP system. I analyzed ERP systems, built maintenance and parts databases, logistics systems, PDM systems, integrated personnel systems, and hooked document/image management in with the enterprise applications- and I learned to do some database design/analysis using both IDEF and CASE/method (ERDs)- but Inever joined the ERP implementers. In the end, data models could only grow to a certain size- although databases could be huge. This limited integration by the big, central database.
Client/Server: As the LAN and PC grew to dominance and integration by a single central database showed both its power and limitations, the client/server model took off as the dominant integration technology. This lasted for several years, maybe until the Internet ramped up. I learned the internals of several databases, some database administration, SQL, ODBC and (later) JDBC administration, VB and the power of the OCX component. I also played around with some other tools of the time. RPC, CORBA, COM, DCOM and COM+: One wave of technologies for integrating across the enterprise involved remote procedure calls. My involvement was never very large or deep, although a bit of this is used in almost any integration effort. I did more analysis than implementation. The reader may find this off, but asynchronous messages predominate in large scale integration. There is a limit on the availability (up-time) of systems composed of other systems via synchronous calls of this kind, found in reliability engineering. The total availability of a system composed of 4 component systems is A=A1 * A2 * A3 * A4. If each was 95% available the combination drops to 81% available.
"If you do anything long enough, you eventually get good at it." MK