Chapter 1. Introduction

Table of Contents

Purpose & Scope

Just about a year ago in May 2000, AEPortal (Advanced Enterprise Portal) started as a typical web project: Put a wrapper around some existing services, retrieve some existing data sources and hide everything behind a common GUI. Clearly the functional requirements were the most important ones at that time.

Based on a rudimentary framework (similiar to an early version of STRUTS from apache.org), modeled roughly after the J2EE architecture the team started to write handlers, models and JSPs.

It looked like AEPortal would become an Enterprise Portal.

In the middle of October last year – when load tests and preparations for deployment started – the non-functional requirements made a sudden and violent appearance and they continue to dominate AEPortal till this very day.

Much has changed since then. AEPortal has been extended in many ways to cope with the non-functional requirements – going far beyond a regular web application. This document describes the portal specific aspects of AEPortal with a special focus on reliability, stability, performance and throughput.

At the end organizational problems that had a major effect on AEPortal are described as well.

Purpose & Scope

As an enterprise portal AEPortal goes beyond a simple web application architecture. It relies on internal multithreading, read-ahead and does a massive amount of caching to concentrate a large number of external services onto a personalized homepage - while still being responsive.

Being a portal AEPortal needs to combine transactional features (e.g. Telebanking) with publishing features (e.g. News). Right now it does a fairly good job on the transactional side but it is very weak with respect to publishing. This needs to change quickly and the document tries to highlight the problems and give hints on possible solutions.

This document also explains the technology used in AEPortal. It discusses critical aspects e.g. in case of external service failures. These topics are not covered in the Infrastructure Architecture Document. (24)

Caching plays a major role in AEPortal. Every site that delivers dynamic AND personalized content has a severe performance problem currently solved or mediated through large hardware investments.

Example 1.1. A little hardware at Charles Schwab

A recently published paper on the portal architecture at Charles Schwab shows that IBM delivered 500! Multiprocessor Unix stations to run the personalized portal. Of course the fact that the portal still runs CGA technology certainly accounts for a larger number of workstations - still - the numbers are frightening both financially and from a system management point of view

Caching of dynamic AND personalized content is both hard to do and an absolute must for such sites. The overall architecture of a high-volume portal is more or less dominated by caching issues. This document collects what AEPortal already does with respect to caching as well as what still needs to be done.

Another goal of this document is to collect ideas from within the AEPortal team or our friends. So please do not hesitate and add to it. There are still a lot of features missing that are vital for an enterprise scale portal (meta-information handling, search features, service development kit, maintainability in production etc.)