I am trying to create a mirrored device with thousand of mapping table entries.
Unfortunately, it seems that the allocated kernel memory to satisfy this configuration is unbearable - I am talking about 6G for 10,000 entries where the used region size is for each entry is minimal (i.e. 8 blocks).
My questions are as follows:
1. Did anyone tried before to create a mirrored device with thousands of targets?
2. What is the estimated total allocated memory per each target?
3. Is there something I am missing here?
Thanks a lot,