Program Analyses for Cloud Computations

摘要: Cloud computing has become an essential part of our infrastructure. In this model, data and programs are hosted in (often third-party) centers that provide APIs for access running large-scale computations. It is used almost all major internet services companies increasingly being considered by other organizations to host their run analytics. However several challenges lie its full-scale adoption, chief among them security, performance correctness. Security important as both client computations need be sent third-party centers. Performance cloud involves development iterations, each on data. Correctness critical frameworks complex distributed systems serving billions users every day.In dissertation, I argue program analysis techniques can help address the above key computing. describe three projects illustrate different aspects solution space: MrCrypt a system uses static guarantee confidentiality using homomorphic encryption schemes. Vega library significantly improves incremental rewriting modified workflows use previously computed results. Kuai distributed, enumerative model checker verifies correctness Software Defined Networks, networking layer many