Add three papers:
1.sinfonia: A new paradigm for constructing an extensible distributed system. This paper is the best paper of SOSP2007, and it expounds a paradigm method for constructing a distributed file system, which is very useful for individuals. Taobao fully referred to this paper when establishing TFS, OceanBase and Tair systems.
2. Fat lock service of loosely coupled distributed system. This paper introduces the distributed lock implementation mechanism Chubby of Google in detail. Chubby is a file-based distributed lock, and Google's Bigtable, Mapreduce and Spanner services are all built on this basis, so Chubby is actually the basis of Google's distributed transactions and has very high reference value. In addition, the famous zookeeper is based on Chubby's open source implementation, but according to friends who work at Google, there is still a gap between zookeeper and Chubby in performance and function.
3.Spanner: Google's global distributed database, which is the first distributed database in the global sense and also a work of Google. Many consistent design considerations are introduced. For simple logic design, atomic clock is also used, which is also of great reference significance in distributed systems.
In addition, there is a book:
Just came out, read the sample chapter, and it feels good. I recommend it to you-large-scale distributed storage system: principle analysis and architecture practice.