I saw a comment a few weeks ago where one fellow HNer ran his entire startup on AWS spot instance pricing - so that he was forced to program in a state of continuous chaos as his demand/spot instances popped randomly and continuously into and out of existence while his service was running. It's like programming on quicksand.
This is probably a step too far - but maybe it is a natural extension of NFLX's Chaos monkey.
If you want ~100% up time with no QOS degradation - your system must be constantly under catastrophic attack.
This is probably a major reason why vol based risk models in finance are completely pointless.
Value at risk of any investable securities (including cash) is 100% all the time - any other number is bullshit. All volatility based risk models are useful if you like watching squiggly lines or pricing options, but they are essentially a random anchor that helps us sleep at night (anchoring bias).
This is probably a step too far - but maybe it is a natural extension of NFLX's Chaos monkey.
If you want ~100% up time with no QOS degradation - your system must be constantly under catastrophic attack.
This is probably a major reason why vol based risk models in finance are completely pointless.
Value at risk of any investable securities (including cash) is 100% all the time - any other number is bullshit. All volatility based risk models are useful if you like watching squiggly lines or pricing options, but they are essentially a random anchor that helps us sleep at night (anchoring bias).