Next 50 Years of Databases

I know that it's been a year since my last post and that I still need to write part 3 in my series on the open research problems for transaction processing database systems. A sneak peak is that (1) we're developing a new distributed DBMS, (2) we're building a large library of "ready-to-run" OLTP applications for testing and benchmarking, and (3) we're creating an on-line encyclopedia of database systems. For these systems, disks were the primary storage location of databases because they were able to store more data than could fit in memory and were less expensive.Although a lot has changed 50 years later in terms of how we use databases, the relational model and SQL are still predominant way to organize a database and interact with it. There are now DBMSs that are designed to ingest new information quickly for on-line transaction processing (OLTP) applications and other DBMSs that are designed to store large amount of data for complex on-line analytical processing (OLAP) programs.These newer DBMSs also take advantage of the three major hardware trends that have emerged in recent years. Beyond obvious things like the volume and velocity of the data being stored being much greater, there will be major changes in how databases are used in applications and the type of hardware that they will deployed on. Thus, the DBMS will store multiple copies of its data in pre-computed materialized views in order to quickly respond to any possible query.The role of humans as database administrators will cease to exist. Again, the tighter coupling between programming frameworks and DBMSs will allow the system to make better decisions on how to organize data, provision resources, and optimize execution than human-generated planning.We will see the rise of database transactions for inter-planetary devices (e.g., space probes).