Because of the higher cost of SSDs, it is especially important to find the right balance between cost and protection against drive failure. Knowing how the performance and availability of Flash Pool aggregates are affected by SSDs becoming unavailable can help you determine the correct approach for your needs.
The failure or unavailability of an SSD affects its RAID group and aggregate the same way as for HDDs. When an SSD being used in a Flash Pool cache becomes unavailable, the RAID group that contains that SSD goes into degraded mode, and the cache performance for that Flash Pool aggregate is reduced until the RAID group can be reconstructed by copying the contents of the unavailable SSD to a spare SSD.
As with HDD RAID groups, if a cache RAID group experiences more concurrent SSD failures than its RAID level can correct for (3 for RAID-DP, 2 for RAID4 ), data integrity and availability can be compromised.
Therefore, for maximum data availability, you should use RAID-DP for both the HDD RAID groups and the cache. In addition, you should have a spare SSD available at all times so that reconstruction can begin immediately in the case of an SSD failure. The spare SSD can be shared between multiple Flash Pool aggregates, although all Flash Pool aggregates using that spare will be without a spare after an SSD fails until the spare can be replaced.
If using RAID-DP and providing a spare for the Flash Pool cache is prohibitively expensive, and if your RAID group size is not larger than the maximum RAID4 RAID group size, the next best alternative is to use RAID4 for the Flash Pool cache (you should still use RAID-DP for the HDD RAID groups), and provide a spare SSD. This is preferable to using RAID-DP with no spare, because double SSD failures are relatively rare, especially with a smaller RAID group size, and reconstruction takes less time than procuring and installing a spare.