Validating and Benchmarking the Availability of an application
This may sound outrageously stupid; however still mustering up courage to throw the discussion out in Open. Is it possible to validate or benchmark availability of the application by running a load test for a certain duration of time?
If yes: then what are the metrics that need to be captured to assure that an application has availability of 99.95% or 99.99% etc.
if no: then why not?
You need to articulate your assumptions and goals. Are you assuming the software is perfect and only needing to quantify the HW failure rate? If so, it’s tedious but tractable.
If your software is going to be expected to survive various failures (e.g. node or link failures) then it’s much trickier. You’ll probably need to run in VMs and simulate various errors (hardware and lower levels of software) and track the fault tolerant behavior. Turning that into formal (credible) statistics is going to be tricky.
Since software interactions are far more common, and harder to predict than hw failure rates … I’m dubious that you can accomplish what you seem to suggest … in any meaningful way.NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!