Possible performance problem with fusion drive

Foxandxss said:

Hello, I am not a guy of moving stuff around but I wanted to format my system so I moved everything to another computer by network.

I ran some problems (finder error 36 AKA I/O) so I started worrying about my disks life. I could say that the problem was the network (I copied some big files around the same drive and no problem at all).

Well, the thing is that I am with my new installation and I started to move my stuff back to the mac and while I was copying the stuff my mac got slow. Chrome tabs that takes a lot longer to show up, apps that usually open in less than a half second now takes 5, the terminal which takes all life to show the prompt... You take the idea.

To be honest, I don't remember if this stuff happened with my old installation (since I don't usually move my files).

I remember the good old days in Linux where you forgot to turn on the DMA in the kernel so the system dies while writing.

I don't think that this is normal and I am not sure what can I try. Ideas?

Thank you and this is my first post here

Click to expand...

Why do you think it is the fusion drive? Jusat wondering.The FDD is made up of two seperate drives, one SSD and one HDD. It should either work or not. There should always be about 5TB available to write on the SSD on average so it should generally run as fast as an SSD for most file access.

However on the initial filling of the drive the FDD can slow right down if you have more files than the size of the SSD portion and spill over into the HDD

In the initial copy of files to an FDD you fill up the SSD first then fill up the HDD (overflow) if you have more files that the size of the SSD. So if you have 500GB of data (OS and files) and the SSD protion of the FDD is say 250GB then you use all of the SSD portion (minus 5GB window for writing fresh files) and the rest goes on the HDD (approx 255Gb). So yes, you can get HDD speeds at the point of 'overflow'.

But once the OS organises the FDD at the block level to move rarely used files to the HDD and keeps regualrly accessed files on the SSD then the FDD should then run similar to an SSD most of the time. The smaller the SSD then you are likely to require access to the HDD, while if the SSD poriton is large then you are likely to fit most regularly used files on it with minimal on the HDD.

Alternately, if you have say a 250GB SSD in a FDD and only have 120GB (for example) of files then you will not use the HDD portion in this example.

If you use iStat Menus you can see the writing/reading on the seperate SSD and HDD on the FDD.

Once you are running in normal mode (everything is loaded and time to settle), and assuming you have say 250GB of data on an Apple FDD (128GB SSD and 1TB HDD) then you will have about 123GB of the SSD filled up with about 5GB spare for bursts of writing fresh files. The OS will keep essential OS files and regulalrly used files on the SSD and rarely used files on the HDD. All this block level movement and 5GB window occurs behind the scenes. Because it is at the block level you could have parts of a file on the SSD and parts on the HDD.

The above summary shows why it is better to have a larger SSD in a FDD so you have more room to have more regulalrly used files.

Hope that makes sense.

Note also my summary is general, and that the FDD is more complicated in operation than I set out.