WIDS 2014 - Part 1
I can mark "attending a conference" off my todo list now. There are an increasing number of monthly dev meetups in Hong Kong, but WIDS (the World Internet Developer Summit) is the only conference. Sure, the name's a bit bombastic and it sounds like a Justice League of developers coming together to save the Internet, but it's a proper conference! Speakers from big name companies! Multiple tracks! A name tag!
The first day was at Cyberport. The previous time I was here was for the Angelhack hackathon, and it's a really beautiful tech startup campus hub thing. When I joined the line, I was pretty sure we were waiting to enter the auditorium where the talks happen. Turns out, the arena we were lining up at is the venue.
I've noticed that Hong Kong events at Cyberport and Science park tend to be preceded by a bit of a report card on how Hong Kong is doing with the startup thing. Speakers from the various organizers came up to remind us about the awesome internet speed here, and the increasing number of startups getting funding from the States.
The first talk was from Microsoft about Azure services. Not that much of interest happened, really. There were brief attempts to link what was being presented to general cloud concepts, but it was pretty much just an hour-long ad. There was a weird five minutes where I think they were just showing us animated graphs on Excel. They completely flew past the only interesting part of the demo, which was when they showed a natural language query interface that allowed you to draw arbitrary analytics by just phrasing it in plain language. The whole thing was like Youtube where you have to wait for the sponsored ad to end before getting to the good stuff.
The next talk was just a complete switch to full on tech mode. Asya Kamsky spoke about replication in MongoDB. I had a vague understanding of replication before, but her talk went into some practical implementation details. A replica set is a collection of servers (usually three) that store separate copies of the same data. One is a primary, and is the one used for both reads and writes in general. The others are secondaries that try to keep up with the state held at the primary. In the event of a failure at the primary, the remaining nodes hold an election and promote one of them to be the new primary. One thing I found cool is that when the old primary does come back, there are mechanisms to allow it to join the cluster and turn into a secondary. The same mechanism allows for downtime-less maintenance. Something I had not thought about before is that replica sets can be useful for running other background tasks such as backups and analytics on one of the secondaries. Overall a really good talk. The one question I had was that these concepts do seem fairly general. What makes MongoDB more suited to this than a relational DB?
Stay tuned for the next epic chapter of the WIDS Chronicles trilogy.