Title: Toward a diagnosis plane for cloud computing
Abstract: Today, debugging is an ad-hoc process. Yet, modern applications running in the cloud (and ones running among clouds) are extremely complex and run on many stack layers (e.g., virtualization, network, host OS). They also rely on many independent parties (e.g., cloud providers, ISPs) that may be reticent to provide sensitive information relevant to diagnosis. This disconnect between ad-hoc debugging practices and the complexity of cloud applications has led to dire consequences: long, costly outages and over 50% of developers’ time spent debugging. In this talk, I will discuss initial steps toward creating a diagnosis plane for cloud computing—i.e., a framework that ties data collected from different information sources together so distributed applications and the ecosystems they operate in can be analyzed holistically. I will also discuss how our past work in inter-domain routing provides insights into how to deploy this diagnosis plane in ways that incentivize all parties to participate.
Bio: Raja is a research scientist at Boston University. His postdoc research focused on ways to evolve inter-domain routing and on creating new routing protocols that can fulfill the needs of today’s modern edge networks (e.g., clouds). His PhD research focused on techniques for automating problem diagnosis in distributed systems and cloud-computing environments.