Yes there are per vdisk and per physical disk limits to the available
resources. For the vdisk its the amount of metadata that can be cached
in memory (called AmapThreshold). The limit for the metadata is around
6% of the available memory (not configurable at the moment but we can
change this). Amap (Address Maps) are metadata blocks which maintain
the LBA mapping, we need this information when we need to read a block
asked by the host. The more these are in memory the better, but we try
to balance the number of amaps per vdisk so that one single vdisk does
not end up caching more there by decreasing the performance of the
others. However if only one vdisk is being used as in this case its
most likely that all the amaps in memory are for that vdisk.
Back to the performance numbers, there are other factors like how much
of the vdisk has valid data. For example the following will lead to
almost no io on the physical disks
1. zeroed data
2. LBA unmapped blocks (blocks which were never written into)
3. Deduplicated blocks which are aleady in a read cache we maintain
> --
> You received this message because you are subscribed to the Google Groups "QUADStor Storage Virtualization" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to
quadstor-vir...@googlegroups.com.
> For more options, visit
https://groups.google.com/d/optout.