From what I understand it's quite different from Microsoft's tiered storage spaces.
With tiered storage spaces, if I have a 600GB SAS HDD and a 500GB SSD, I'd have a 1.1TB disk with some stored in the SSD and some in the HDD, based on usage patterns. This is (a) nice because you get to use all the space, and (b) a bit scary because if anything fails, you'll lose everything.
With Virtual SAN, it looks as if the L2 cache is pure cache, so if I have a 600GB HDD and a 500GB SSD, I'd have a 600GB disk with lots of cache. As I understand it, if it's a write-through cache, then all the data is stored on HDD before the writer thinks it has been written. In this case if the SSD fails, I won't have lost any data. So, I don't need to have two SSDs in RAID 1, because loss of the SSD won't cause any loss of data, so I may as well have 1TB of RAID 0 L2 cache (not on the 600GB drive, that'd be silly
). (If the L2 cache is set as write-back, then I will probably still want RAID 1)Is the above correct?
I'm not sure if the disk will still be accessible automatically if the L2 cache device fails. Even if not, is it simply a case of reconfiguring the device to not have any L2 cache, and I'll be able to access the data (more slowly)?
Is there any documentation on how the L2 cache works - ie is it MRU, LRU, LFU, etc? Does it try to keep stuff in the L2 cache longer than it would otherwise to try to save write cycles? (Some caches will cache all reads, and the algorithm just decides which old stuff to throw away - that'd be bad for an SSD cache, so you'd need a clever algorithm to decide what to actually cache in the first place).
