“Valar Morghulis” is the tenth and final episode of the second season of Game of Thrones. Valar Morghulis is a common greeting in Braavos, meaning “all men must die” in High Valyrian. It is meant in the sense of “all men must (eventually) die,” sooner or later.
Creating a fictitious parallel, it is plausible to say that your data will (eventually) die. To be more precise, when using Single Failure Tolerance (1FT, also known as FTT1 or RF2) there is an extremely high data-loss probability of 0.49% in one year or 1.95% over the 4-year life of a system with a with once-a-month disk scrubbing. Less frequent scrubbing increases the data-loss probability to as high as 18.6% in the fourth year of a system’s life.
It is a myth that 1FT, a configuration commonly quoted by hyperconverged vendors, can handle one disk failure.
You don’t need to take my word for it, but instead, read the study “Single Failure Tolerance (1FT): Myth and Fact” proposed by Lakshmi N. Bairavasundaram, Zhe Wang, and R. Hugo Patterson. The trio has enough credentials and credibility, to make you think twice before using 1FT for your HCI or Converged platform deployment. Download the technical paper Here.
This article was first published by Andre Leibovici (@andreleibovici) at myvirtualcloud.net