Use both SSDs and HDDs in pool, only use HDDs as duplicate/secondary?

[TomWaitsForNoMan]

Hi there,

I currently have a single pool made up of HDDs with 2x duplication. I want to buy a few SSDs and move to a structure where:

1) There is one set of data on the SSDs and a duplicate on the HDDs

2) Any new files are written to the SSDs before being duplicated to the HDDs

3) Any file reads are read off the SSDs

 

So effectively the HDDs would be a duplicate of the SSDs but not the primary source of data. Is this possible to set up in Drivepool, either in a single pool or using hierarchical pools? If not exactly, then even approximately would do.  Thanks

[Shane]

I think you'd need to use hierarchical pools. Something fairly simple like:

Pool A = Pool B + Pool C where Pool B = all SSD and Pool C = all HDD; Pool A duplication x2 but not real-time; Pools B and C duplication off.

DrivePool should be able to detect that B is all SSDs and prefer it for reading and writing, and duplication happens nightly (as per settings.json file; default 2am) when not real-time. Note that "prefer" doesn't mean "guaranteed" however.

But if you wanted to get a bit more clever, you could have something like:

Pool A = Pool B + Pool C where Pool B = all SSD and Pool C = one SSD + rest HDD; Pool A duplication x2 real-time; Pools B and C duplication off; Pool C have SSD Optimiser enabled with the SSD as the cache and the Automatic balancing set to empty the cache to the archive HDDs nightly (or whatever).

DrivePool would then be duplicating in real-time but still at SSD speeds, at least unless/until you filled the cache SSD (which would empty nightly or whatever), it should still prefer the all SSD pool B for reads, and even if it read from the pool C with HDDs it might still be reading (new files) from that pool's SSD cache.

[YiZixuan]

@Shane 

I believe your solution is not user-friendly enough. In fact, the SSD Optimiser doesn't seem to be a user-friendly design either, in my opinion.

I am looking forward to a feature for SSD caching that requires user actions:

1. Add SSD1 to the pool.
2. Allow the user to designate SSD1 as a cache volume (or allocate a portion of SSD1's storage space for caching).
3. The user can choose to persistently cache a specified `folder1` in the cache space or enable automatic caching.
4. The user can opt to enable features for `prefer reading from cache` and `prefer writing to cache`.
5. The user should be able to manually clear the cache space or configure a cache expiration time.

Then, everything else should be handled automatically by the software, such as:
1. Gradually loading the contents of the user-specified `folder1` into the cache space during idle HDD I/O.
2. Reading from the cache space when there is a cache hit.
3. Automatically loading files into the cache for future use when the user reads them.
4. Releasing cache space for files that have not been read within the cache expiration time.
5. Cached files will be written to the pool.

Most importantly, any files in the cache space should not count towards the backup quota; they should only serve as cache.

This is the functionality I envision.

[Christopher (Drashna)]

Unfortunately, the SSD Optimizer isn't a true caching solution.  And it's not meant to be.  

That said, StableBit CloudDrive operates in a way that is more like a caching solution, and in line with your expectations.  But again, it's not a true caching solution, either. 

It sounds like you want a full on caching solution, like PrimoCache