April 5, 2024

A paper reading club at work; databases and distributed systems research

I started a paper reading club this week at work, focused on databases and distributed systems research. I posted in a general channel about the premise and asked if anyone was interested. I clarified the intended topic and that discussions would be asynchronous over email, run fortnightly.

Weekly seemed way too often. A month seemed way too infrequent. Fortnightly seemed decent.

I was nervous to do this because I've been here about 2 months. In the past I would have waited 6 months or a year to do this. But I don't know. If you see something you think should exist, why wait?

The only other consideration was past experiences I've written about having difficulty getting engagement with clubs at work. But EDB has near 1,000 employees. I figured there might at least be a couple interested.

Furthermore I figured if I only got a few people this entire idea would at least benefit myself, since I have been wanting to force myself to build a paper reading habit. And if no one responded, it would be only mildly embarassing and I'd not pursue it further.

But after a day, about 6 people showed interest. Which was better than I hoped! Folks from product management, support, development, and beyond.

So I opened a dedicated channel and asked people to start submitting papers and voting on them. One of my teammates started submitting some great papers on caches and reference counting.

I picked a first one, the Redshift paper, to get us started. Demonstrating the process to avoid confusion. And I made a calendar invite for everyone in the channel, the paper linked in the invite. I clarified in the invite that it was just a reminder and that the real discussion would still be async over email. (I've found it's best to repeatedly clarify process stuff.)

Once I had these first few folks interested I was able to post again in a broader company channel that a couple of us were starting this paper club. By the end of the day the dedicated channel was 29 folks. All in about 2 days.

Mailing lists are nicer than Slack or Discord in my opinion because they sort of force you to slow down, they are harder to miss (if someone starts a thread after you've seen a message in Slack or Discord, you tend to miss it), and easier to manage (read/unread).

Engineers often seem to get overwhelmed by a mass of Slack messages. Whereas they seem to be a bit more comfortable with email threads.

All of this is all the more important when you're running a global group. EDB has people everywhere.

Why do this?

Before I dropped out of college I did a research internship with a VLSI group at Harvard SEAS. And my favorite part was that they had a weekly (or biweekly?) Wednesday paper reading session where 15 people from the lab and adjacent labs would eat pizza after hours and discuss a paper.

I've been dying to recreate this at a company ever since. Since EDB is so distributed, we won't be discussing over pizza. But I'm still excited.

And I hope my experience serves as a blueprint for others.