JDBC? We Don't Need No Stinkin' JDBC: How LinkedIn Scaled with memcached, SOA, and a Bit of SQL
TS-4696

Have you built your site around the JDBC API and MySQL database or Oracle Database, only to find it slowing when you need it the most? Have you found that more customers means more hardware and more sleepless nights? No, this isn't an infomercial for yet another SOA boondoggle. This session shows how LinkedIn and many other high-scaling Web sites are storing their most precious data and, even more importantly, how they keep to ACID rules while still responding to user requests from external caches. Higher user loads mean more opportunity to interact and sell to your users but also mean that you need to have the systems to respond to their requests. Come learn about how to scale large back-end systems that stay ACID from the end user's perspective but scale with open-source technologies, to many cheap machines, without using the dreaded two-phase commit.

The session provides

• A quick overview of JDBC and caching technologies
• Gap analysis of most systems out there today
• An explanation of how LinkedIn uses open-source technologies (memcached, MySQL database, Tomcat, Jetty, and Java technology) to build a scalable data storage tier
• Best practices for storing user-generated content in multiple languages and in a way that allows for more languages and features -- without rebuilding the tables and/or the file structure
• Best practices, including phased conversion and rollout -- no need for a massive all-or-nothing conversion