When you’re adding servers and data centers as fast as Facebook, standardization and automation are your best friends. At the Open Compute Summit IV, held in Santa Clara in January, Facebook’s Delfina Eberly provided an overview of how the company uses standardization and automation to manage data centers at Internet scale. Eberly, the Director of U.S. Data Center Operations for Facebook, said the company has effectively automated all repairs that didn’t require hands-on attention. As its growth accelerated, Facebook constantly assessed and updated its tools and workflow, and developed an integrated spare parts portal so inventory stocking and parts replenishment was built into the workflow. Facebook also developed a custom ticketing system specific to its needs. While the company is known for its freewheeling engineering culture, data center operrations is a different matter. This is one place where we put a lot of rigor into how we operate,” said Eberly. “In the data center, we put a lot of rigidity into the workflow.” This has reduced the need for “tribal knowledge” in the data center, said Eberly which was important as the company rapidly added new data center techs at various locations. This video runs about 30 minutes.
For more new about Facebook’s data centers, see the Facebook Data Center FAQ or our Facebook Channel. For additional video, check out our DCK video archive and the Data Center Videos channel on YouTube.