On the resilience of mission critical applications
Justin Y. Shi · 2015
Most mission critical applications are distributed onto multiple computers. They need to survive extreme conditions. Today, regardless methodologies, all existing mission critical distributed applications suffer the "performance or reliability" scaling dilemma. Each such application will crash on any transient or permanent software or hardware failure. These failures cause losses in data, time, money and human lives. In 2012, a computer architecture inflection point was declared.This paper reports that the root cause of this phenomenon is the use “virtual circuit” as the basic building block for program-to-program communication. Existing communication protocol implementations do not support the application level data transmission reliability as promised by the virtual circuit concept. In fact, any transient or permanent software/hardware failure in the data packet assembly layers, such as TCP/IP (Fig. 1), can break the application level virtual circuits causing the crash of the entire application [2].