About technological barriers

PlugDB is a full-fledged DBMS (Database Management System) embedded in secure hardware. PlugDB stores data in relational tables, provides indexing and SQL query capabilities, ensures the logical (integrity constraints) and physical (transactional atomicity) data integrity, and most importantly, protects data through declarative access control policies (i.e., which user can access the data satisfying a given SQL predicate).

The implementation of such an embedded DBMS raises many scientific challenges related to the combined constraints of the microcontroller (e.g., very little RAM) and NAND Flash (e.g., cost of random writes, Block-erase-before-page-rewrite, wear leveling, etc.). Tackling these challenges requires a profound redefinition of the classical data management principles (storage, indexing, querying, transactions).

In particular, an efficient query evaluation with little RAM requires to massively index the database. The consequence is to generate many random writes in Flash to maintain the indexes, which has an exorbitant cost. Known solutions to reduce random writes in Flash consume a significant amount of RAM, thus leading to a vicious circle. To break this vicious circle, PlugDB offers new techniques to structure the database and to evaluate queries.

Furthermore, the execution of distributed protocols on a large number of personal servers poses new challenges. For example, how to perform an epidemiological study over tens of thousands of distributed medical records in as many personal servers as number of patients? How to process statistical consumption calculations on millions of smart meters scattered among dwellers without re-centralizing the data? PlugDB addresses these challenges by defining distributed query protocols on an architecture called asymmetric. On the one hand, this architecture is composed of a large number of personal servers that are low powerful and weakly connected but highly secure, and on the other hand, of an external infrastructure (e.g., the cloud) that is very powerful, highly available, but without confidentiality guarantees.

Next: Healthcare