Seems a rather vague question right?
It can depend on any number of variables:
- Tyre type
- Tyre condition
- Age of the driver
I know starting a tech blog post off by talking about the stopping distance of a car is odd but I wanted to make a point as this is relates to a question I was asked by a customer recently.
A fairly snr figure in my customer organisation asked me in a public forum “How long does it take to power on a Vblock”. With the benefit of hindsight (isn’t that a wonderful thing), I should not have even entertained trying to answer his question in that forum. But of course I tried, with both parties dissatisfied of the quality of my answer.
As I tend not to give disappointing answers to questions this has troubled me for a few weeks now. Analysing my response over and over didnt seem to resolve my issues with my answer so I have come to the following conclusion
Fundamentally the question was floored!
For example is he referring to how long it takes to boot up the individual components after power outage? Is he referring to getting the base level virtual infrastructure back up (likes of vSphere etc)? Or is he (as i suspect) referring to how long it takes the application to be back to a state where they can run their services?
Each view point has a plethora of facets that need to be factored in to the SLA variables, like below just to mention a few:
- MTTR – Mean Time To Recovery
- RTO – Recovery Time Objective
- RPO – Recovery Point Objective
In a mixed workload environment (which this is) it is common to see different workloads having different SLAs within the same system, and depending on the logical setup around each of those workloads (cluster settings, SRM, Stretched DC clusters etc) each SLA can vary wildly.
Thus, “How long does it take to Power on a Vblock” is not only an impossible question to answer (without a multitude of variables to support the questions) but its fundamentally the wrong question as each application will take differing levels of time to restore fully.