1.1 Motivation

Computers have been increasingly influencing our lives in the last decades. Distinct eras of computing have elapsed, driven by technological progress and affecting the way we are using computers. These shifts of paradigms were motivated mostly by advances in hardware and software technology, but also by changes in the way humans interact with computers.

The first computers were large machines solely built for dedicated applications such as numeric computations. There was no real distinction between software and hardware, and new applications could not be executed without changing the physical configuration. The instructions were executed directly without any operating systems underneath and the program had direct access to all system resources. Also, there was no interaction between the running program and the user. With the advent of punch cards and batch processing, computers became more flexible tools for processing jobs. Not just data--executable programs, too--were now used as input, defined in low-level, imperative languages. There was still no interaction during execution and jobs could only be executed sequentially.

Machines were expensive but in great demand. The next innovations were thus influenced by concepts allowing multiple users to work on the same machine and multiple programs to run at the same time--mainframe computers, operating systems, and time-sharing. Terminal-based command line interfaces provided the first interactive systems. New programming languages such as FORTRAN or ALGOL allowed the development of larger and more reliable applications without using pure assembly code.

The arrival of networking was also the beginning of a new kind of computing. Distributed applications not just take advantage of the resources of multiple machines, they also allow the existence of federated systems that consist of multiple machines remotely located. It promptly became apparent that in order to support such systems, additional software on top of a network operating system was necessary. Others favored a distributed operating system which provided a high degree of transparency and supported migration functions as part of the operating system. Early middleware systems based on transaction monitors and remote procedure calls emerged, easing the development of distributed applications. Existing networks were linked and a global network was formed; the Internet. The evolution from mainframes to mini computers, then workstations and personal computers, not just urged networking. It also introduced new user interaction mechanisms. Consoles were replaced by graphical displays and enabled more sophisticated forms of interaction such as direct manipulation.

At that time, the idea of object-oriented programming arose and quickly gained much attention. New programming languages such as Simula and Smalltalk emerged and embraced this principle. After a short time, objects were also considered as distributable units for distributed middleware systems.

The following era of desktop computing was shaped by personal computers, growing microprocessor clock speeds, graphical user interfaces and desktop applications. At the same time, the old idea of non-linear information networks was rediscovered and its first real implementation appeared. Dating back to Vannevar Bush's MEMEX device and later Ted Nelson's concept of documents interconnected through hypertext, the WWW represented a truly novel distributed information architecture in the Internet.

With the first commercial usage, the WWW scored an instant success and soon hit a critical mass of users to become the most popular service within the whole Internet. This also motivated a complete transformation of our information economy. Technological advances at that time predicted the rise of mobile computing that should eventually proceed into an era of ubiquitous computing. Computers were not only going to become smaller and smaller, but also omnipresent and produced in various sizes and shapes. This introduced the notion of calm technology- -technology that immerses into everyday life. Ubiquitous computing also heavily changes the way we are interacting with computers. Multi-modal and implicit interactions are favored and provide more natural interactions. Devices such as laptops, tablet computers, pad computers, mobile phones and smartphones have already blurred the boundaries between different types of computing devices. Combined with wireless broadband internet access, they have introduced new forms of mobility, providing connectivity at all times.

While we are currently on the verge of ubiquitous computing, there are other trends as well that are influencing the way we think about and do computing right now. The progress of microprocessors has saturated in terms of clock cycles due to physical constraints. Instead, modern CPUs are equipped with increasing numbers of cores. This trend has forced developers, architects and language designers to leverage multi-core architectures. The web has already started to oust desktop applications. Modern browsers are becoming the new operating systems, providing the basic runtime environment for web-based applications and unifying various platforms. The web is changing and provides continuously more user-centric services. Being able to handle and process huge numbers of users is common for social platforms such as Facebook and Twitter and corporations like Amazon or Google. Hosting such applications challenges traditional architectures and infrastructures. Labeled as so-called "Cloud Computing", highly available, commercial architectures emerged. They are built on large clusters of commodity hardware and enable applications to scale with varying demand over time on a pay-per-use model.

Now let us take a step back and summarize a few important developments:

  1. The web is a dominant and ubiquitous computing technology. It will replace many traditional desktop application environments and provide ubiquitous information access.
  2. Web applications have to cope with increasing demand and scale to larger user bases. Applications that incorporate features such as collaboration and web real-time interaction are facing new challenges as compared to traditional web applications.
  3. Multi-core and multiprocessor architectures will dominate the processor landscape. At least for the next decades, performance gains of processors will mostly be attributed to increasing numbers of cores or processors and not to increased clock cycle frequencies.
  4. Large, scalable systems can only be designed when taking into account the essence of distributed and concurrent systems. Appropriate programming languages, frameworks and libraries are necessary to implement such systems.

In this thesis, we will bring together these individual developments to have a comprehensive analysis of a particular challenge: How can we tackle concurrency when programming scalable web architectures?