Pavel Klinov, Stardog VP of Research and Development, is back from the Dagstuhl Seminar on Graph Databases and Network Visualization, held at the Leibniz Center for Informatics in Germany from January 16 – 21, 2022. We asked him about his experience.
Q: What makes the Dagstuhl Seminar different from other events?
Dagstuhl is very different from the traditional conference, where most time is spent listening to talks with occasional panels, workshops, and networking events. Dagstuhl is a relatively small gathering, 20-25 people, who split into even smaller breakout groups and try to figure out novel ways to approach open problems in the field.
I am new to this format. It’s not easy to get an invitation to Dagstuhl, and I never was invited during my academic years. Honestly, this invitation came out of the blue last Spring. I never really made any contribution in the visualization area. But it was fine; many visualization folks had little exposure to graph databases but were keen to listen and learn, so I did the same.
Q: How was the seminar useful to a database engineer who doesn’t write research papers?
This was the question I kept asking myself in the run-up to the seminar. It’s a means where the end is building better products.
Q: How did Dagstuhl help that?
In several ways. First, as surprising as it was for me, graph visualization and graph query optimization may have some interesting overlap. Talking with visualization people made me realize that many visual layout methods require graph analysis and summarization techniques (like frequent pattern mining, etc.) that sound eerily similar to what the database system does to better understand the data for optimizing queries. The whole thread of “understanding the data” ran big throughout the seminar, and it surely feels like research engineers in both areas have things to learn from each other.
Q: How else?
Query result representations in many database systems, Stardog included, have room for improvement when it comes to visualization. The possibilities go well beyond various charts that work for tables. One example is path query results. Those aren’t tables, and they can easily explode in dense graphs, thus overwhelming users. This problem has been known to network visualization researchers for decades, so it’s only natural to look into the various compact representations and visual layouts that have been developed over the years.
Q: Final thoughts?
Finally, of course, there is no downside to making Stardog better known in the research community, particularly the database community. We are always looking for talented engineers to work on cutting-edge graph database tech. It may well happen that some of the industry-leaning students or researchers would join our engineering force in the near to medium future. If designing next-gen query optimization algorithms appeals to you, don’t hesitate to ping me today!