Storage spaces two way mirror

[frankr2994]

Ok of someone could explain a few things to me. I've been testing and reading alot but something's remain a mystery. I have 4x500GB 2.5 5400rpm drives for testing. I've been playing with setting them up as a two way mirrored storage space. First question is when I set up all the drives and then I loaded about 500GB of data onto the pool. I then removed one of the drives and it evenly split the data on the three remaining drives. I thought this was odd so I sent over another 24GB file that again was dispersed 3 ways. How did my two way mirror become a three way mirror? Benchmarks shown with 4 drives write speeds where around 80MBps and 3 dropped it to 40. Second question is since storage spaces should stripe data why does my 4 drive array test at 80 when I have any single drive test at 80. But if I set up two mirrored pools and then stripe them using disk management the new array test at over 140. I have no plans on combining storage space mirrors and disk management stripes it was only a test. Ultimate goal is to have 4x4TB drives set up as a two way mirror using refs file system with 2 virtual hard drives inside. I just want them to run as efficient as possible. I'm using windows 10 pro and building the array in a windows 10 workstation vm so I can use refs.

[jaclaz]

I believe you might be confusing the number of copies (or amount of redundancy) which is 2 in a two way mirror, with the number of drives in the drive pool (3 or 4 in your case).

This latter, from what I understand, is not connected with the former.

Basically your (say) 500 GB of data is first "divided" into 256 MB units called "slabs", which are "doubled" (i.e. "mirrored") and then "dispersed" on the disks that part of the pool.

You do not have anymore 500 GB of data, but rather some 512,000/256=2,000 * 2 = 4,000 "slabs" of which "roughly" are written 1,000 on each of the 4 disks.

When you remove one disk, the whole stuff is (or should be) re-assembled making a new copy of each slab that was on the removed disk, and again dispersing them on the available disks, so each of the three disks should have about 1,333 slabs.

Compare with:

https://techcommunity.microsoft.com/t5/storage-at-microsoft/deep-dive-the-storage-pool-in-storage-spaces-direct/ba-p/425959

jaclaz 

Edited January 18, 2020 by jaclaz

[frankr2994]

Well what confuses me is it being called a two way mirror. Meaning in a 2 disk setup each disk should have an identical copy of 100 percent of the data. What I didn't realize is that I guess in my 4 drive setup is that it only gives me one column of data. I could manually configure 2 columns using power she'll when building the pool but the minimum number of drives that I would need to add to the pool would be 4 to keep it efficient. With 2 columns it would basically be a software raid 10. When this server needs upgraded the most I can see being spent would be for 2 more drives at a time so I'll leave it at the default. I am assuming the drives I had in mind which are 4tb wd gold Enterprise drives should read/write fast enough to flood the gig Ethernet connection which in reality is all I can ask for. I may start the build with 5 drives and leave the 5th raw and not part of the pool so I can easily add it in if a drive is showing signs of failure.if I can get my hands on a spare SSD I may play with Intel rst to make it a cache drive. I'm not interested in SSD tiering with storage spaces because I only need read cache and should only need one drive for it. At any rate I have about 2 weeks to decide how I want the storage before it needs to be put into use.