Most enterprise portals lack performance and or stability during the first couple of releases (e.g. Deutsche Bank, yellowworld etc.) and AEPortal was no exception. Current project methodology was not helpful here because it did not recognize the different needs of an enterprise portal:
There is NO proven knowledge about how to build a large-scale enterprise portal
Most of the current technologies are IMMATURE (Java, Web Application servers etc.) bug ridden or have low performance and stability (Java1.1 on multiprocessors etc.)
Load-tests are no longer simple acceptance tests (spend a couple of weeks and put the checkmark behind the milestone). The load-tests are needed to MAKE the portal perform – in other words: they will cause a cycle of engineering and software changes over weeks and months. There is no chance of an enterprise portal to support an open user group right from the beginning.
The load tests do not only cause software changes. In many cases they force the enterprise to re-engineer some of its backend services because they do not scale in the context of the portal. Project management needs to manage and track this process.
The end-to-end environment is extremely complex and CANNOT be tested in one go with a huge enterprise portal application. It is absolutely nonsense to start portal testing BEFORE every component in the whole processing chain (from load-balancing and reverse proxies web server, web application server(s), databases and backend service connections) has been tested INDIVIDUALLY.
