Open Source Bridge 2011: Day 1 Liveblog

1630: Rather than hitting the night’s parties and socials, I’m going back to my room to chill. Have to pace myself. 🙂

1545: Massively Scaling Django for a Global Audience with Playdoh, Frederic Wenzel. The speaker works at Mozilla, which has more than 140 live websites. Lots of sites, lots of load, lots of scaling. About 35 web developers divided into seven teams to support this. This doesn’t include QA or IT. “Mozilla loves Django.” Playdoh = Django plus libraries plus docs. Playdoh is a grab-bag of low-level libraries.

Mozilla add-ons web site = 5 MM page views/day. Firefox input = 30 K page views/day. The smallest server setup: 1 load balancer, 2 web servers, 2 MySQL, 2 memcached servers. Performance, optimization, localization, security, and deployment tools. The Playdoh library suite looks very cool, I’ll keep it in mind for IP Street.

1430: A dozen databases in 45 minutes, Eric Redmond. In trying to make this topic more approachable, the speaker gives short shrift to ACID, BASE, the CAP theorem, and a host of other topics. This is a bad survey talk. The speaker seems very smart, so, maybe he’s just ill-prepared for his talk today.

1330: Queues, Eric Day of Rackspace. Great overview of queuing technology. AMQP, ZeroMQ, etc.

1200: Lunch was a cold Mediterranean buffet spread, and it was perfect.

1100: Intro to CouchDB, J. Chris Anderson. Primary focus of this talk is using CouchDB on mobile hardware. Always funny to hear a less-than-well informed young person give his view of computer & Internet history and tradeoff; gee, I never realized that computers were boring before the Internet.

Using CouchDB replication to create shadow db’s between a master db and the users clients, to create and enforce information security. Nice architectural model. An excellent talk, although it focused more on mobile applications than I had expected.

1000: Cloud Scaling, Gavin McQuillan. AWS is first, Rackspace is second for Ycombinator startup server hosting selection. EBS volumes suck, they aren’t fast, only a couple of hundred iops/sec per volume. EBS also competes with network traffic, including Ephemeral Disks. You get only ~ 10 GB of true local storage on any EC2 instance. Ephemeral Drives are faster than EBS, but go away if a node is terminated. Intended for scratch space. “Anything super I/O-intensive isn’t going to scale well in the cloud.”

0850: In the Sanctuary waiting for the Keynote to begin. Nice crowd here. Android users outnumber iPhone users by a wide margin. 🙂

Updated as the day unfolds…