DIMA Lecture by Lukas Schwerdtfeger

The Internet of Things (IoT) presents a novel computing architecture for data management: Databases and Stream Processing Engines (SPE) need to perform query processing in distributed, heterogeneous, mobile, and potentially hostile environments. To avoid expensive dynamic lookups for query and schema information during query execution, databases and SPEs use query compilation to generate optimized machine code.

Traditionally, SPEs depend on an operating system such as Linux to provide networking and memory management primitives. This makes them impractical for use in small IoT devices. Additionally, conventional query compilation methods only address the query execution and not the I/O operations, where the operating system is a boundary that cannot be optimized.

Unikernels eliminate the boundary between the application and the operating system, which makes them an ideal candidate for low-latency network applications. Unikernels allow aggressive query compilation across syscall boundaries by configuring typical runtime parameters at compile time. In addition, unikernels have a fast boot time, a small footprint, and promise a high degree of security, making them an excellent fit for the IoT.

Lukas Schwerdtfeger is a student at the Database Systems and Information Management research group (DIMA) led by Prof. Dr. Volker Markl. He will demonstrate a prototype implementation that reuses components of an existing stream processing engine, NebulaStream, to create query-specific unikernels that enable reliable low-latency stream processing compared to the traditional approach.

More information on NebulaStream: https://nebula.stream/