Autoren: Gregor Hohpe
Google is known to operate one of the largest civilian computing infrastructures. These hardware resources are managed by a vast collection of software frameworks and tools, which form the basis for highly parallelized, reliable, low-latency, high-throughput applications. They also provide useful programming abstractions that speed up development and debugging. Some parts of this infrastructure, such as MapReduce, GFS, Sawzall, Chubby, Protocol Buffers, are available as open source projects or published in academic papers, while others are proprietary. Rather than dive into the dark corners of each of these tools, this talk tries to distill key design themes and patterns, which enable these unique capabilities, and can be re-used in other contexts.
Prerequisites: a basich understanding of distributed programming