My PhD thesis is about the design, the implementation, and the evaluation of data management solutions to build persistent applications, that is, applications that can seamlessly execute on computing systems despite unexpected crash or machine shutdowns. The goal is to devise algorithms that optimize the trade-off between consistency criteria and performance for hybrid servers featuring NVDIMM and DRAM.
My work specifically focuses on heavily multi-threaded applications, that is, applications that can leverage large multi-core processors to achieve high performance. To optimize performance, I study the influence of the synchronization mechanisms between threads and the impact of data placement strategies.
EuroSys'22: Ana Khorguani, Thomas Ropars, and Noel De Palma. "ResPCT: fast checkpointing in non-volatile memory for multi-threaded applications." Proceedings of the Seventeenth European Conference on Computer Systems, pp. 525-540. 2022.
Compas 2021 : Ana Khorguani, Thomas Ropars, and Noel De Palma. ”Trade-off between performance and
transparency for in-NVMM checkpointing.” French-speaking computer science conference in Parallelism, Architecture and System. 2021.