The traditional function of minimizing slow disk accesseses in the absence of reasonable sized caches is largely gone as an interesting research area today. You may want to look at ways of storing data in memory and/or distributed systems generally. For example, asset data storage and retrieval in games, succinct data structures, systems like etcd.
Another tangential advice: look at some of the recent published research. Eg ACM SIGMOD proceedings.
The landscape tends to have relatively shortlived fads that leave behind them a narrow tail of applications that are actual fits for the idea in question. (Industry examples: "big data analytics" -> Hadoop, "web scale online db" -> MongoDB, etc).
YottaDB is a modern company proving an M db.