Massivizing Distributed Systems = Making Systems Scalable, Reliable, High-Performance, etc., Yet Efficient

Invited Talk, Lorentz Center, the Netherlands

Download PDF Slides


Computer systems are present in every aspect of our lives. From business, where online shopping is ICT-based, to education, increasingly online and digital, to science, increasingly using ICT as scientific instrument, to leisure, where online gaming is the most popular form of entertainment. This has all become possible only in the new networked world, where networking advances enable everyone to access very sophisticated computation. With companies such as Google, Facebook, Amazon already present in our daily routines, is there anything still needed from ICT that the industry cannot yet provide? The answer is an emphatic "Yes!" Although we can already develop computer systems, we have grasped the main laws of their operation, and we have developed theories about how computers work, at massive scales computer systems behave differently from what we expect. This lecture focuses on interesting new challenges in the operation of the computer systems, grouped under the umbrella-term of Massivizing Computer Systems. The challenges are many: modern computer systems under-perform, become unavailable, scale poorly, are less elastic than we wish, work as complex software stacks about which we cannot reason well, form complex systems of systems whose ecosystem-like behavior we are just beginning to grasp. You will learn about early approaches to understand the laws of modern computer systems, and about our early theories on the design, development, deployment, analysis, and benchmarking of computer systems that are massive in nature.