Given the ubiquity of cloud computing on the one hand and the (well-hidden) complexity of providing elastically scaling distributed services on the other, one wonders how cloud architects can describe such scenarios. Remembering the heydays of object-oriented programming, seasoned readers might even hope for something similar to the unified modeling language (UML) for cloud services.
This excellent survey, however, completely shatters any such hope. Searching for service description languages (SDLs) that cover the whole life cycle of a cloud service--from modeling to deployment, including service discovery and service-level agreements (SLAs)--the authors find 67 different SDLs for the 2008 to 2019 time range. Most SDLs focus on service deployment and provisioning, with aspects of service discovery and service selection formalized to a very limited extent. Due to widely varying focus areas, SDLs are typically not interoperable (with the notable exception of TOSCA defined by OASIS).
There is also no trace of a comprehensive and holistic single description language for every aspect and stakeholder of cloud services, like UML does for generic software engineering. The authors suggest the unified service description language (USDL), but a superficial search failed to uncover strong continuing support after the World Wide Web Consortium (W3C) completed early work in 2011. The research also identifies that aspects of the Internet of Things (IoT) are typically out of scope of almost all SDLs to date.
I very much like the paper’s comprehensive comparison (2.5 pages) and short characterizations of every single SDL, following a four-stage life cycle categorization. As this is the best thing one can do in such a heterogeneous environment, I unconditionally recommend the paper to every cloud architect looking for a suitable modeling language. The only aspects one could add to this otherwise excellent survey are indicators of who is actually driving the standardization (easy to check) and to what extent the SDLs are really used. But that is possibly a research paper on its own.