Non-volatile random access memory (NVRAM) is an emerging storage technology unifying the
advantages of DRAM and classical block-based secondary storage devices such as HDD and SSD,
and is expected to replace the latter in database systems in the near future. Existing works have
shown that especially hybrid-memory systems smartly combining DRAM and NVRAM are a
promising architecture. However, NVRAM is subject to specific errors such as NVDIMM failures
necessitating data replication to prevent primary data loss. Current state-of-the-art systems apply
a synchronous physical replication of the persistent data to the NVDIMMs of remote sockets
in a scale-up system by using a master- slave approach, whereby readers can only access the
master. This approach yields a high overhead, but has no benefit beyond fault tolerance. The goal
of this diploma thesis is to investigate, in how far a novel structurally polymorphic replication
representing different replicas of the same base data in different data structures and using them
to facilitate efficient query processing can alleviate these drawbacks.