Drive balancing?

SDWolf

I've noticed some odd behavior with DrivePool's balancing, and I think the mechanism could use some refinement.  While writing to the drive that has the most free space sounds like a good idea, this presents a problem when the pool consists of drives of varying sizes.

Imagine a pool consisting of two 3TB drives, two 2TB drives, and two 1TB drives.  Currently, DrivePool won't write anything to the 2TB drives until the 3TB drives are at least one-third full, and will keep those 1TB drives empty until the 3TB are two-thirds full and the 2TB drives are half-full.

A more balanced approach would be to write to the drives that have the most percentage free space, rather than absolute free space, to more evenly spread the data (not to mention the I/O load and wear & tear) across all of the drives.  I would rather have all of my drives 30% full than have all my data concentrated on just my largest drives.

lee1978

Hi

I have asked a similar question for instance if you where add a new drive I would like the option to populate it with data already in the pool useful for general new drive or upgrades. Also I would like the option to spread the data of 1 drive to the rest of the pool for instance if a drive was about to fail or needs upgradeing would be nice feature

For instance I am about to add another 8tb of storage to my server 4 x2 tb drives my problem is these drives will only take the new data I add

Lee

Alex

I will talk about the motivation behind the current balancing algorithm. Balancing is actually a pretty complicated math problem when you start considering duplicated files. M3 solves this, and we'll probably be tweaking it.

The motivation for the current balancing strategy is to avoid as much balancing as possible and to only balance when we are sure that moving a file will benefit the utilization of your free space.

However, I do recognize that some people might not share the same goal so there will be other options down the road for balancing strategies.

mike1812

Thanks for the explanation Alex.  I've been running DrivePool for  a few weeks now and I too have noticed the balancing "issue" above.  I understand the current logic, but do look forward to options down the road, as I agree with the above posters - would much rather realign my drives to 30% across the board, for example. 

edit: nevermind the rest I had originally posted here; I found the answer on the blog - thank Goodness - the forum posts were confusing me!

Robert

Personally, I don't care how the files are organised on my drives as the original WHS used to take the approach of filling up one disk before moving on to the next. There are performance advantages to spreading the data across as many disks as possible of course, but I'm happy with the current solution implemented by Alex because my main reasons for using DrivePool are:

1. Being able to have a share that is bigger than any one disk in my system (e.g. having a share with 2Tb of files in it, even though I might only have 3 x 1 Tb drives).
2. Being able to protect my data via duplication.