Am Montag, 26.10.2020 hält Prof. Dr. Stefanie Scherzinger (Universität Passau) einen Vortrag im Rahmen des Informatik-Kolloquiums.

Der Titel lautet: „Pain Points with NoSQL Data Stores“.
Der Vortrag findet in zoom statt.


During the last decade, we have seen radical changes in the way software is being built: Release cycles have accelerated from yearly to weekly (if not daily) deployments. In the spirit of lean development, design decisions are made as late as possible. This also holds for the database schema: Fields that might be needed in the future are not presently added to records, reasoning that until the next release, requirements might change in a way that would render these elds unnecessary after all. It is partly due to this need for more exibility that NoSQL data stores such as MongoDB, CouchDB, and Google Cloud Datastore have become so very popular. As they are schema-free, the schema is often only implicitly declared by the application code.

Yet NoSQL data stores bear their own pitfalls when it comes to writing scalable and maintainable software. This talk discusses four actual pain points, where the rst three are related to agile schema evolution, such as
1. application runtime errors and even data loss, when schema changes are carried out lazily and in an uncontrolled fashion,
2. the person hours wasted when developers write throw-away code for eager data migration, out of lack of a declarative data denition language,
3. the diculty of assessing the technical and monetary migration debt when working with multi-versioned, persisted data,
4. and nally, how data access patterns where hot spot data objects manifest threaten application performance.

Along with each pain point, I sketch solutions that target these problems. I also present insights from our empirical studies of NoSQL schema evolution in open source database applications, conducted within the DFG-funded project \NoSQL Schema Evolution and Big Data Migration at Scale“.