London, England, United Kingdom
SDE @ AWS Athena
SDE @ Prime Video
As a researcher at EPFL, I study the interference between concurrent transactional (OLTP) and analytical (OLAP) database workloads. Real-time analytics are very important for companies that need to make time-critical decisions. At the same time, SLAs signed between service providers and clients impose restrictions on the performance of the aforementioned workloads. My goal is to provide real-time analytics running OLTP and OLAP workloads in the same system on the same copy of the data providing performance isolation by assigning the execution of the two workloads in different processing units, i.e. a CPU and a GPU.
During my PostDoc I research efficient ways to store and process arrays inside the MonetDB database. Elements in arrays, in contrast to elements in relational tables, have a natural order. This is a limitation and an advantage at the same time. Elements cannot be processed out of order, or should be put back in order before being projected but the order of the elements is known in advance and positional lookups can be completed in O(1). In practice, some meta-information is needed to allow positional lookups. The challenge is to minimise this information and develop efficient algorithms for processing it.
Improve and extend the geo-spatial module of MonetDB