I was thinking of upgrading everything but I was reading another thread were someone just upgrade their cpu/memory off ebay and it was much cheaper and gave a big boost. So, I can get a CPU upgrade for $170 and double the memory for $40 would it give me a decent boost if I'm running Plex and Windows VMs? The big difference I see is the 8 threads vs the 4, I'm assuming Unraid can take full advantage of this for the VMs?

 

http://cpuboss.com/cpus/Intel-Core-i7-2600-vs-Intel-Core-i5-2400

 

 

Certainly upgrading to a CPU with more cores will give you more cores to create VM's with. You can also run plex in a docker which is more efficient and does not require an entire virtual machine for. Also, depending on what you are doing with your virtual machine, they usually don't require more then 2 virtual cores, unless you are doing something very cpu intensive, but for gaming, 2 cores is fine. So if you have a dual core i5, you have four cores to work with and can easily setup a VM with two cores running windows, if you want more VM's then certainly upgrading to a quad core i7 will help.

Certainly upgrading to a CPU with more cores will give you more cores to create VM's with.

 

Note that upgrading from an i5-2400 to an i7-2600 is NOT adding more cores ... they're both quad-core CPU's.  The difference is that the i7's cores are hyperthreaded, which makes task-switching between threads much more efficient because of the extra register sets; but is nowhere near the impact of actual additional cores.    Nevertheless, the i7 IS a higher-end CPU with about 40% higher performance than the i5  (PassMark 8268 vs. 5840) ... so it will definitely provide better performance for your VM's.

 

HOWEVER ... before doing the CPU upgrade, I'd double your memory => this alone may give you a big enough performance boost that you'll decide you don't need the higher-end CPU.

 

Also, as ashman70 noted, if you're using a VM for applications that are available within a Docker, using the Docker may boost your performance as well, since you'll reduce the overhead associated with that application (e.g. Plex).

 

I doubled the ram to 16GB and got the CPU upgrade. I feel this will get me another couple of years out of the box. I also switched the last plugin to docker last night so everything (important) is a docker.

 

Should I worry about moving the dockers/plex/vms to the cache disk, would that give me some more speed too?

Unless your cache disk is an SSD, you likely won't see any performance increase by moving them.

Is that because writes to the array are going through the cache drive anyway? So moving it I'm not really saving much?

 

I'm recording video in my Windows VM (security cameras) and when I turn them on to recording 24/7 the load spikes up from what I assumed was because it was writing to the array constantly. Oh, shit, I just through of something...maybe my vms are not using the cache disk. That would be bad right?

The cache drive is not part of the array, so it is not affected by reads/writes to the array. I may be wrong with my assumption that moving things to the cache drive won't improve performance, it might. Maybe someone with more knowledge can chime in on this.

The cache drive is not part of the array, so it is not affected by reads/writes to the array. I may be wrong with my assumption that moving things to the cache drive won't improve performance, it might. Maybe someone with more knowledge can chime in on this.

 

When you write to the array parity is calculated for each bit, right? So, if you write to a non array drive (cache drive) parity isn't calculated therefor it should be faster. I think that's the general purpose of the cache drive, since parity isn't written it's much faster to write to it.

Should I worry about moving the dockers/plex/vms to the cache disk, would that give me some more speed too?

I would recommend moving your dockers and VMs to the cache disk (or another disk outside the array).  There shouldn't be much overhead when a docker/vm loads/reads from the array, though it certainly will be faster from an SSD just because the SSD is faster than a spinner.  Writes are a different story, though.  When you write something to the parity protected array it isn't just an I/O operation - parity needs to be calculated.  Therefore, if your dockers/VMs are constantly writing to temporary files or similar then it would be expensive to constantly be doing those writes against the array.  Is this an issue for you?  I don't know, it depends on what your dockers/VMs are doing.  Still, I think it's a better practice to locate dockers/VMs on the cache drive/pool given the possibility of high I/O operations against the docker/VM image.

 

In particular, it's important to move the Plex /transcode location to the cache drive (or RAM).  Plex isn't just using your CPU to transcode, it's writing the transcoded stream to disk and then streaming what has been written.  If Plex is writing the transcoded data to the array then you have a lot of unnecessary overhead.

 

I prefer htop to top for assessing CPU load.  It's more graphical (in a text based way) so it's more difficult to cut and paste, but it will show you all the cores instead of just a made up load average.

You'll absolutely speed things up a good bit if you set the VM's to use the cache drive -- this will be at full disk speed instead of the much slower writes to the parity protected array.

 

Is is easy to move VMs/dockers?  My Windows VM is constantly writing security video to disk so I'm assuming that's hammering my cpu/disk since it's on the array.  I should move that off asap.