Brownouts are an Issue, But Might be Almost Unavoidable

Brownouts tend to be a typical feature of most networks using internet protocol.  Where most measures of availability (reliability, we sometimes call it) measure times or percentages of times when a resource is unavailable to use, brownouts represent the times or percentage of times when a network or resource does not operate at designed levels of availability.

Just as an electrical brownout implies a severe drop in voltage but might not be an outage, a network brownout follows a sharp degradation in link quality but might result in the affected circuits still being technically “up,” Oracle says. “This decline may be triggered by congestion across the network or a problem on the service provider’s end.”

Brownouts are in one sense “a feature not a bug,” a deliberate design choice that prioritizes resiliency over guaranteed throughput. That is the whole architectural principle behind internet protocol, which sacrifices routing control and quality of service on defined routes in favor of resiliency gained by allowing packets to travel any available route. 

And since the availability of any complex system is the combined performance of all cumulative potential element failures, it should not come as a surprise that a complete end-to-end consumer user experience is not “five nines,” though enterprise networks with more control of transport networks and end points might be able to replicate five nines levels of performance. 

The theoretical availability of any network  is computed as 100 percent minus the product of the component failure rates (100 percent minus availability). For example, if a system uses just two independent components, each with an availability of 99.9 percent, the resulting system availability is less than 99.8 percent. 

Component	Availability
Web	85%
Application	90%
Database	99.9%
DNS	98%
Firewall	85%
Switch	99%
Data Center	99.99%
ISP	95%

source: IP Carrier 

Consider a 24×7 e-commerce site with lots of single points of failure. Note that no single part of the whole delivery chain has availability of  more than 99.99 percent, and some portions have availability as low as 85 percent.

The expected availability of the site would be 85%*90%*99.9%*98%*85%*99%*99.99%*95%, or  59.87 percent. Keep in mind that we also have to factor in device availability, operating system availability, electrical power availability and premises router availability. 

In choosing “best effort” over “quality of service,” network architects opt for “robustness” over “reliability.” 

Source: Digital Daniels