When we first started capturing realtime analytics data for Kernl it was a simple affair: 1 table N columns (id, date, product_uuid, domain, product_version, php_version,….) This worked well for the first few weeks, but eventually we started to see serious performance degradation as the table size grew. The first step we took was to normalize… Continue reading Migrating Kernl Analytics from Batch-Driven to Realtime