On a last-minute ditch I submitted a short abstract to the ICTOPEN 2018 event - an annual event organized by the Netherlands Organisation for Scientific Research (NWO). Based on this abstract I was invited to present a poster about my research on performance challenges in serverless computing.
With the poster I wanted to highlight the two main facets of my research: (1) the research done as part of my personal research, such as the Fission Workflows prototype; (2) the joint work on serverless computing with the SPEC RG Cloud research group, highlighting our current work on a reference architecture for Function-as-a-Service Platforms.
As my first poster session, this was an interesting and slightly challenging experience. Although I prepared an extensive pitch beforehand, having to pitch this emerging research domain to outsiders revealed what is not clear about the domain to outsiders, and clarified the current knowledge of the average researcher on serverless computing. Aside from that, for the next session I will need to reduce the pitch length even further - it was too long according to the feedback I received.
Abstract - As a key part of the emerging serverless computing paradigm, Function-as-a-Service (FaaS) platforms allow users to run arbitrary functions without being concerned about operational issues. However, there are several (performance-related) challenges surrounding the state-of-the-art FaaS platforms that can hinder widespread adoption of FaaS, including size-able overheads, unreliable performance, and the absence of benchmarks. In this work provide an overview of our ongoing work towards addressing these challenges. Thus far, we have been building a workflow engine, Fission Workflows, to provide efficient function composition within the FaaS model. Furthermore, together with the SPEC CLOUD RG, we have been working on solving more high-level challenges; defining a common terminology and reference architecture, as well as developing a industry-wide benchmark of these FaaS platforms. Besides presenting our work so far, we also propose a roadmap towards solving these challenges to invite researchers to join this effort.
https://t.co/osAz3UfGbj 2018 has started: @erwinvaneyk explains the #serverless work of @LargeResearch to Dick Epema of @EEMCS_TUD. (Very proud moment: Dick trained Erwin's supervisor, and his interest now help train Erwin.) pic.twitter.com/5XZ9L5jaeH— Alexandru Iosup (@AIosup) March 19, 2018