Life lessons and datacenter performance analysis

Amer Diwan · 2014

Summary form only given. Performance is critical to datacenter applications: a poorly performing application provides poor experience to end users and also wastes hardware resources (network, disk, CPU). Unfortunately, these applications are huge (serving each user request can involve hundreds to thousands of machines) and complex (many interacting components often developed and managed by different teams). Thus, these applications are notoriously hard to understand and optimize. This talk distills key insights from our experience in analyzing and optimizing datacenter applications (particularly Gmail) at Google.

Read the paper · More papers on PaperTik