Celeron G1610T maxed out from simple file transfer to cache.

[Mr_Who]

Hi guys.

 

First of, I'd like to say that it might just be me who are wrong about what I am about to ask.

 

Is it normal for UNriad to utilize 87% procent of a dual core Celeron G1610T, maxing out the transfer speed at at around 90MB/s, from client to SSD cache on the UNriad server? It just seem so odd that this task of transferen a large single 22GB file to a SSD at something like 112MB/s is an impossible task to ask a Celeron G1610T server?

It's only the cache working and I can confirm the drive/drives don't spin up for the client to server transfer. I just don't understand how this task can max out the CPU?

 

In fact, I feel like the newest version: 6.11.5, have made everything slow? I have a Qnap TS-453A which also suddenly can't seem to perform the same numbers as it used to. I can't prove this since I have no logged numbers to back this up. It just seem like things are going heavy on the systems.

 

I bought this HP Microserver Gen8 with a Celeron G1610T CPU and 4Gb of RAM for a senior friend of mine. The idea was to run it as a NAS with UNraid and a simple Jellyfin server, so he had a large central storage unit for all the films and movies he wants to digitally backup.

I was the one selling him on the idea of using this server as a setup due to it's fair price listed on the used market.

Now I am not sure if the server will even be able to keep up with the job of running Jellyfin without any transcoding...

 

 

Am I setting my expectations too high for this old hardware?

[Frank1940]

So run some more tests and see if you get the same results.  Make sure you don't have any other parity operations doing on. See what the process utilization is when you are not copying files to the cache.  Look to see what disks are having read/write operation occurring. 

 

The Celeron G1610T is considered to a low midrange CPU (Passmark rating).  Is it that much more work to setup a Jellyfin Docker and see what happens?  (I would think transcoding is out unless you can assign a GPU to the task. Assuming Jellyfin supports that...)  A screen capture of CPU section and the Docker Container section of the Dashboard tab showing the condition might be helpful.

 

Attach a copy of your diagnostics file (captured after you have seen the issue) in a NEW post in this thread.   (Other wise no one will know that you have uploaded them.) 

Edited January 26, 2023 by Frank1940

[Kilrah]

Where are you looking at CPU usage? The Unraid dashboard includes iowait which will make it seem higher than pure compute load when doing transfers, especially with few cores. Compare the reading with htop in the terminal, it'll likely be way lower there. 

 

But yeah, that was already a low tier CPU when it came out a decade ago.

Edited January 26, 2023 by Kilrah

[Mr_Who]

22 minutes ago, Frank1940 said:

So run some more tests and see if you get the same results

Do you mean setting up different versions of UNraid?

 

23 minutes ago, Frank1940 said:

The Celeron G1610T is considered to a low midrange CPU (Passmark rating).

I know. Yet, slower CPUs have been used to obtain full utilization of the gigabit port.

 

25 minutes ago, Frank1940 said:

Attach a copy of your diagnostics file (captured after you have seen the issue) in a NEW post in this thread.   (Other wise no one will know that you have uploaded them.) 

I will try this. The problem is a constant, not a periodical one.

[Mr_Who]

23 minutes ago, Kilrah said:

Where are you looking at CPU usage?

I just looked at the dashboard readings. Regardless of what the actual utilization of the CPU might be, I can see the transfer from the client to the cache is slow, which I do not understand. How can transfer to an SSD cache without the parity calculation be bottlenecked somehow?

[Frank1940]

You will be waiting a long time if you are waiting for a response from someone who is still running a Celeron G1610T.  To begin with, this is a ten-year old processor.  We are trying to help you figure out if that processor will be able to do the job for your friend.  It makes no difference if the processor is running continuously at 100% IF it can still do the job.  (By the way, looking at the GUI consumes CPU cycles also so that has be to factored in when evaluating low end CPU's.)  The requested information (diagnostics and screen captures) is needed to look at the system and see what might be impacting the system-- if anything.   The questions asked were to make sure that (1) you were not running a parity check (as an example) in the background or (2) it was a one time observation (yes, that has happened!).

 

As another observation, If that processor load is running at 98% for 70% of the time and doing the job flawlessly, you have the perfect setup for the job at hand!  Furthermore, I suspect that if it is not up to the task, you can find another LGA1155 socket processor (perhaps, on E-bay) for a very reasonable price to replace it.

Edited January 26, 2023 by Frank1940