Home > Storage > PowerMax and VMAX > Data Protection > Dell PowerMax and VMAX All Flash: SRDF/Metro Overview and Best Practices > Understanding bias
As described previously, bias is an integral function of the two VMAX3 or VMAX All Flash arrays taking part in a SRDF/Metro configuration. The initial createpair operation places an SRDF device pair into an SRDF/Metro configuration and pre-configures the bias to the primary or R1 side of the device pair by default. From then on, the bias side is always represented within management interfaces, such as Solutions Enabler SYMCLI or Unisphere for VMAX, as the R1 and the non-bias side as the R2.
In the case of a failure causing the device pairs to become not ready (NR) on the link, SRDF/Metro responds by making the non-biased or R2 paired device inaccessible (not ready) to the host or host cluster. Bias can optionally be changed by the user once all SRDF device pairs in the SRDF group have reached ActiveActive or ActiveBias SRDF pair states. As noted previously, changing the bias to the R2 side effectively swaps the SRDF personalities of the two sides with the original R2 device pairs now being represented as the R1. Changing bias to the R1 side would be redundant as the R1 personality always follows the biased side.
In both examples above, a failure has caused the SRDF/Metro device pairs to become not ready (NR) on the link, which resulted in the biased or R1 side remaining accessible (read/write) and the R2 or non-biased side becoming not ready (NR) to the host or host cluster. The left example represents a single host configuration with the default bias location after a user initiated suspend operation, while the right example depicts the resulting post failure configuration after a change in bias was made.
As noted previously, there are failure scenarios for which bias alone would not result in the ideal outcome for continued host accessibility. In the example below, a failure affecting the R1 or biased side would result in both the R1 and R2 (non-biased) sides becoming inaccessible to the host or cluster. For these scenarios, the optional and highly recommended redundant witness protection provides the best host accessibility outcome.