Thursday, October 3, 2019

Pat O'Neil thoughts

The relatively small database kernel community has lost another one of the Old Guard: I recently learned that Pat O'Neil passed away last month.

I was lucky enough to have Pat as a colleague, although briefly. My first "real job" after I graduated from college back in the mid 1980's was at Computer Corporation of America in Cambridge, MA, and I sat just down the hall from Pat and followed his work closely as he led the team building support for B+ Tree indices for Model 204.

This was the first time I learned about "the ubiquitous B-Tree." But, more importantly, it was being around Pat with his enthusiastic fascination for database kernels and file structures and access methods that launched me on my path to 40 years of being a storage systems engineer, the best career on earth in my opinion!

I left Boston for California, but 10 years later I met up with Pat again, when he invited me up to Seattle to learn about the new SQL Server team that Microsoft were setting up. Due to a variety of reasons, I didn't pursue the SQL Server opportunity, choosing instead to go work at Sybase, but I really appreciated the fact that Pat thought of me and took the time to introduce me to some fascinating engineers at Microsoft.

In addition to being brilliant, Pat was extremely nice and was a wonderful role model for a young engineer like me.

Moreover, he was extremely lucky in another way: his wife, Betty (I of course knew her as Professor O'Neil) was just as brilliant, and was a great teacher. As a fledgling graduate student at the University of Massachusetts, Boston, I took her graduate class in Operating Systems and it was a tremendous experience for me. My undergraduate work was all in pure Mathematics (Chicago didn't even have a Computer Science department at the time!), and I had no idea what a wonder a well-taught Computer Science class could be like.

Pat is, of course, best known for inventing the LSM-Tree, which is now the mainstay of many a modern DBMS, but his work on the LRU-K buffer management algorithm, his work on the C-Store, and the marvelous "A critique of ANSI Isolation Levels" are all just as powerful and just as timeless.

I didn't realize it until I looked at his Wikipedia page, but not only did Pat and I both study (decades apart) at the University of Chicago, but the odds are reasonably high that Pat was a student of my father's back in the early 1960's, at MIT. I guess perhaps Pat and I were even more connected than I knew.

I'm sorry you're gone, Pat. I'll miss you.

No comments:

Post a Comment