Preserving Recomputability of Results from Big Data Transformation Workflows
The ability to recompute results from raw data at any time is important for data-driven companies to ensure data stability and to selectively incorporate new data into an already delivered data product. However, data transformation processes are heterogeneous and it is possible that manual work of domain experts is part of the process to create a deliverable data product. Domain experts and their work are expensive and time consuming, a recomputation process needs the ability of automatically adding former human interactions. It becomes even more challenging when external systems are used or data changes over time. In this paper, we propose a system architecture which ensures recomputability of results from big data transformation workflows on internal and external systems by using distributed key-value data stores. Furthermore, the system architecture will contain the possibility of incorporating human interactions of former data transformation processes. We will describe how our approach significantly relieves external systems and at the same time increases the performance of the big data transformation workflows.
KeywordsBigData Recomputability System architecture Bitemporality Time-to-consistency
- 1.Accumulo A (2017) Accumulo design – data model. http://accumulo.apache.org/1.8/accumulo_user_manual.html#_accumulo_design. Accessed 27 July 2017Google Scholar
- 4.Dragojević A, Narayanan D, Nightingale EB, Renzelmann M, Shamis A, Badam A, Castro M (2015) No compromises: distributed transactions with consistency, availability, and performance. In: Proceedings 25th Symposium on Operating Systems Principles, ACM, pp 54–70Google Scholar
- 9.Lee J, Muehle M, May N, Faerber F, Sikka V, Plattner H, Krueger J, Grund M (2013) High-performance transaction processing in sap hana. IEEE Data Eng Bull 36(2):28–33Google Scholar
- 11.Rahm E, Do HH (2000) Data cleaning: problems and current approaches. IEEE Data Eng Bull 23(4):3–13Google Scholar