NoWog: A Workload Generator for Database Performance Benchmarking
Parinaz Ameri, Nico Schlitter, J. -P. Meyer, Achim Streit · 2016
The evaluation and benchmarking of NoSQL databases is challenging due to the variety of query languages and also due to the denormalized scheme that allows to include non-scalar values and repeatable structures such as nested documents. In the rather young age of NoSQL databases, there are not many studies on a conventional solution to enable a fair comparison of their performance with each other and with relational databases. In this paper, we present the modeling and evaluation of an extensible open-source solution called NoWog. Its purpose is to generate large synthetic database workloads and also to mimic real-world database applications allowing application-specific benchmarking. We introduce a generic language for the description of workload characteristics that is independent of a particular database query language or the underlying storage model and therefore applicable to various database management systems. Thereby, the NoWog language covers different distributions of read and write operations for various data types such as numbers, texts, arrays and nested documents. In our evaluation, we demonstrate NoWog's ability to mimic a real workload from an application using environmental satellite data stored in a MongoDB and also its scalability to produce big workloads.