Towards High-performance and Trusted Cloud DBMSs

Abstract Cloud Database Management Systems (DBMSs), such as cloud-native analytical or serverless DBs, are experiencing rapid growth in adoption due to their flexibility and scalability. However, recent incidents with cloud providers show that the traditional model of a trusted provider/admin no longer applies to protect the customers’ data. One promising solution that can prevent a sole reliance on cloud and database service providers are trusted execution environments (TEEs). While past TEEs had many limitations and caused high performance overheads, recent work shows that the support of TEEs like Intel SGX for DBMS workloads improved significantly. Thus, it is time to actively integrate TEE technologies into cloud DBMSs to achieve better security that does not rely on the cloud provider. In this paper, we discuss directions for how recent TEEs can be used to build efficient and secure databases. We summarize the recent results on Intel SGX’s performance for DBMS workloads and lay out the remaining research challenges that must be addressed to achieve optimal performance and thus minimize the performance cost for additional security.