Kingsum Chow, Alibaba System Software Hardware Co-Optimization
When you are tackling many servers in the data center, saving a small percentage of servers would bring a significant return. We will describe how we identify performance opportunities at scale, and also how it is different from optimization on a single system.
The emergence of large-scale software deployments in the data center has led to several challenges: (1) measuring software performance in the data center, and (2) optimizing software for resource management. We will share a few techniques we developed to tackle some of these challenges. We will also cover several important considerations to avoid making mistakes in performance data analysis.
The key workshop takeaways are:
- Avoiding common mistakes in interpreting system performance data
- Interpreting CPU utilization and capacity correctly
- Drawing the correct conclusion by discovering Simpson’s paradox in the data sets
- Considerations in making decisions that scale