Heading off correlated failures through independence-as-a-service

摘要: Today's systems pervasively rely on redundancy to ensure reliability. In complex multi-layered hardware/software stacks, however - especially in the clouds where many independent businesses deploy interacting services common infrastructure seemingly may share deep, hidden dependencies, undermining efforts and introducing unanticipated correlated failures. Complementing existing post-failure forensics, we propose Independence-as-a-Service (or INDaaS), an architecture audit independence of redundant proactively, thus avoiding INDaaS first utilizes pluggable dependency acquisition modules collect structural information (including network, hardware, software dependencies) from a variety sources. With this information, then quantifies interest using auditing modules, offering various performance, precision, data secrecy tradeoffs. While most general efficient assume auditor is able obtain all required can employ private set intersection cardinality protocols quantify even across unwilling their full with anyone. We evaluate practicality three case studies via realistic structures.