When you can no longer provide quality manually
In his slideshow "Achieving QoS in Complex Distributed Systems through Autonomic Computing," Dr. Daniel A. Menascé (Dept. of Computer Science at George Mason University) lays out two approaches to Quality-of-Service management:
- Overprovisioning (expensive and still can be overwhelming if peak load surpasses system capacity).
- Dynamic reconfiguration features (to adapt to changes in workload intensity).
With dynamic reconfiguration, you can still achieve excellent QoS from existing capacity. The author also covers design of an autonomic ecommerce system and the challenges establishing QoS can bring about. He then covers easy-to-understand motivations for autonomic computing systems:
- Modern computer systems are complex and composed of multiple tiers.
- The workload presents short-term variations with high peak-to-average ratios.
- Many software and hardware parameters influence system performance so that manual reconfiguration isn't an option.