Vast Design Space

A half-century of database systems research yielded numerous data system architectures, each optimized for a specific set of applications. Designed along multiple dimensions, such as data layouts, storage architectures or recovery strategies, application architects and software developers are faced with a plethora of different feature sets and design options to choose from. This vast design space is still growing as changes in hardware and applications introduce new concerns that warrant new techniques.

Design space

Custom-tailored Designs vs. Limited Resources

Today, matching a scientific or commercial application with its perfect data system is a time-consuming task that not only requires expertise in the area of databases, but also a willingness to compromise. Often, off-the-shelf solutions will only provide suboptimal performance. However, building a custom-tailored system for the task at hand is an expensive endeavor. Modifying an existing system under today’s monolithic implementations is extremely complex, while designing and building a new data system from scratch requires expertise and tens of man-years worth of time.

Difficult Choices

Self-designing Data Systems

Rather than chasing changes in workload and hardware by continually designing and implementing new systems from scratch, or forcing end-users to settle for suboptimal solutions, we envision self-designing data systems that smoothly and autonomously navigate the design space to quickly generate the optimal solution for a given application. Self-designing data systems would relieve both system designers and end-users of data management headaches, culminating in greater productivity. Moreover, a self-designing system may discover new architectures that researchers would have never even considered by synthesizing new solutions out of existing ones, mimicking the natural process data system architects are performing manually.

Solution: Self-Designing Data Systems

Publications

2015
  1. Stratos Idreos Data Systems That Are Easy to Design. ACM SIGMOD Blog, 2015.
  2. Sam Xi, Oreoluwa Babarinsa, Manos Athanassoulis, Stratos Idreos Beyond the Wall: Near-Data Processing for Databases In Proceedings of the International Workshop on Data Management on New Hardware (DaMoN), 2015.

People

Sponsors