The document discusses SQL Server 2012's new indirect checkpoint algorithm. The current checkpoint algorithm flushes all dirty buffers during a checkpoint, causing unpredictable recovery times and IO spikes. The indirect checkpoint algorithm calculates a new minimum recovery LSN during checkpoints rather than flushing buffers. It uses a background recovery writer thread to flush pages when the dirty page count exceeds a threshold, maintaining predictable recovery times without checkpoint IO spikes. The presenter advocates testing and monitoring the new algorithm's performance benefits when enabling it on databases requiring guaranteed recovery time objectives.