In our latest IBM #vSurround discussion, our panel discusses which workloads are not well suited for a serverless environment.
Daniel Krook, Senior Software Engineer, IBM (@DanielKrook)
– David Wells, Software Engineer, Serverless Framework (@DavidWells)
– Adam Johnson, Co-Founder & CEO, IOpipe (@adjohn)
– Ryan Scott Brown, Senior Software Engineer, Red Hat (@ryan_sb)
David: If you are running applications or data-analyses for longer than 5 minutes, it may not be useful to run serverless.
Ryan: You should check the costs of what works most efficiently for what you are trying to do. If you are running big data, Hadoop may be the answer and there may be other frameworks that are more suited for specific tasks. You need to consider your costs to migrate versus the advantages of running in serverless.
Adam: Low-latency workloads – if you have a flat line of traffic vs. “spikey” or less reliable scale timing for traffic patterns, it wouldn’t make sense to go serverless. It is a matter of understanding how to build an application on serverless even when there may not be a lot of documentation so there is a need for users describing how they built their serverless applications.
IBM Owner: Kevin Allen