The last day started early: 5am. San Francisco has a superb transit planner (or perhaps Toronto is the only big city that does not have one) and we went to see the Golden Gate Bridge lit by the sunrise. How could we not? Thus, below is the billionth picture of the Golden Gate Bridge.
Friday had the highest collision rate for good talks – it was tough to choose. One track of interest was the ‘Architectures you’ve always wondered about’ – candid presentations about how Digg and Facebook are solving scaling. One common word: memcached. Later, a presentation with impressing numbers by Randy Shoup, eBay Distinguished Architect: their automated search & recommendation engine processes 50PB daily, with 50TB added every day (no typos :-). 10% of their inventory is in flux every day, a rate of change that imposes challenges when users require accurate search results. As the presenter put it, ‘our users do go to the very last page of search results and they do complain if their number differs by one from what was originally stated’. Whereas with Google, the total number of searches for “Java” is about 414 million, but you cannot see the results beyond the first thousand. Speaking of Google, though, they’ve just announced their new record in sorting 1 PB of data.
I caught the final part of Eric Evans’ Strategic Design talk – it was a packed room, he’s the man behind the “Domain-Driven Design” and, perhaps even more famously, the assertion that “the de-facto standard software architecture is … the Big Ball of Mud”. This is how you can tell he’s worked in the trenches. Check out an identically-titled presentation recorded earlier.
Another track on the last day covered non-relational distributed document storage: Neo4J, CouchDB, Hypertable et al. I went to the second, Jin to the third. Stuff useful to be aware of. If you need this kind of storage you already know more about them.
Some final thoughts:
The conference is overall quite good for its focus on the architecture and high scalability. It’s more than the sheer scale itself: it’s the ingenuity seen in the presentations or on the hallways. Some good stuff should come out of it.
The agile track overlaps somehow with other conferences, yet it’s a sign that how software is developed cannot be separated from the software itself. QCon is a pretty young conference, but one with promise. If you want numbers, my estimation based on how many tables were in the lunch room is somewhere in between 300-350 people. The main organizers from InfoQ were around every day and you could chat with them – the event does not have a ‘mega’ feel to it.