0

I've been thinking about defining some infrastructure-independent metrics for SLA requested by customer.

The developed software is being deployed on-premise within customer's DC and managed by customer's technical staff, - therefore I cannot give any numerical reliability guarantees related to infrastructure (Availability/Fault-tolerance), - this must be stated in the sales agreement.

Another point here - is that both availability and FT form not just of infrastructure and operational aspects, but also of the software quality (it's architecture), which leads to the question -- is it feasible if I defined infrastructure-independent availability and FT somehow (just for developed software in conjunction with underlying applications, like k8s, OS, DBs etc.)? If so, is it ok to use same metrics (availability %%; RPO; RTO; something else?)?

Perhaps there are some books or articles describing my case?

impulsgraw
  • 887
  • 3
  • 13
  • 34
  • How could you imaged the available of a software without hardware? A software is not running in vacuum... Even if you could have written a software with six nines if it runs on a two nines infra then at the end the availability is two nines. – Peter Csala Jul 06 '22 at 14:55
  • @PeterCsala, the software itself supports availability/ft. For example Kubernetes mitigates software faults (e.g. simple unexpected segmentation fault) by re-running the pod and supporting the number of pods within deployment constant. Another example is the hypervisor automatically rebooting the VM with guru meditation. All these take some time, which produces concrete RTO values taken over probability of fault over an hour over a year. Etc etc etc. **Note: I didn't mention a word about physical infrastructure.** – impulsgraw Jul 12 '22 at 21:04
  • These supervisions are done not by your software itself rather some 3rd party tool. So, which components should be included in your estimate and which should be excluded? – Peter Csala Jul 12 '22 at 21:27
  • @PeterCsala just as stated in the question, **is it feasible if I defined infrastructure-independent availability and FT just for developed software in conjunction with underlying applications, like k8s, OS, DBs etc.** I could implement the kubernetes-like system as part of my application, if you want. Anything which is not hardware. – impulsgraw Jul 12 '22 at 21:39
  • @PeterCsala consider [this](https://stackoverflow.com/questions/2483873/software-fault-tolerance) for ex. The only thing I'm curious here is numerical metrics I could use, not just words. – impulsgraw Jul 12 '22 at 21:42

0 Answers0