Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


What is a good load for server?
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

What is a good load for server?

Hi,
The question not as server admin, it's from reseller customer sight.
When I connect to WHM I can see server load at the top right corner,
What can be good load for the server? (I mean how I can avoid overloaded webhosting)

Regards.

Comments

  • What? Are you racing for discussion count?

    Thanked by 2Privacy bulbasaur
  • @chocolateshirt said:
    What? Are you racing for discussion count?

    Really not.

  • The ideal load is a thick, creamy one.

  • mywebhostingmywebhosting Member, Host Rep

    @ChaoscripT said:
    Hi,
    The question not as server admin, it's from reseller customer sight.
    When I connect to WHM I can see server load at the top right corner,
    What can be good load for the server? (I mean how I can avoid overloaded webhosting)

    Regards.

    If you see the load average of 0 which means a server with 1 core CPU is on the idle state. But if the server is running a single process with running wit 100% it will show 1.0 in load average. Similarly, if you have a 6 core CPU server and all the 6 cores are running processes with 100% then it will show the load average of 6.0.

  • Depends on how many cores the server has.

    Why don't you share this and also the load you see, we can then advise.

  • @mywebhosting said:

    @ChaoscripT said:
    Hi,
    The question not as server admin, it's from reseller customer sight.
    When I connect to WHM I can see server load at the top right corner,
    What can be good load for the server? (I mean how I can avoid overloaded webhosting)

    Regards.

    If you see the load average of 0 which means a server with 1 core CPU is on the idle state. But if the server is running a single process with running wit 100% it will show 1.0 in load average. Similarly, if you have a 6 core CPU server and all the 6 cores are running processes with 100% then it will show the load average of 6.0.

    This. The load indicator also takes in account like I/O wait or weird kernel struggles. It is not a good indicator imo. Better measurement is to look at actual cpu usage and ram allocation and have headroom for both like 25% depending on the quality you want to offer.

    The best load is a happy load

  • @noobjockeys said:
    Depends on how many cores the server has.

    Why don't you share this and also the load you see, we can then advise.

    Here is results from "top -c" command in Terminal,

    top - 08:48:36 up 12 days, 20:24, 0 users, load average: 5.29, 7.93, 8.11
    Tasks: 2 total, 1 running, 1 sleeping, 0 stopped, 0 zombie
    %Cpu(s): 13.8 us, 4.7 sy, 1.5 ni, 78.8 id, 0.4 wa, 0.0 hi, 0.7 si, 0.1 st
    MiB Mem : 64150.5 total, 7317.2 free, 17062.4 used, 39770.9 buff/cache
    MiB Swap: 4096.0 total, 2632.0 free, 1464.0 used. 31502.6 avail Mem

    Regards.

  • @ChaoscripT said:
    What can be good load for the server? (I mean how I can avoid overloaded webhosting)

    tl;dl: no more than the number of cores you have, preferably a fair bit less most of the time. Short peaks higher are usually fine, longer or very regular peaks are a sign of trouble.


    It is a noticeably more complicated than that one measure unfortunately, but where possible you want it to be below the number of cores you have. Preferably more than a little under so there is spare capacity should a sudden glut of new requests come in.

    The figure presented for “load” is usually the 1-minute average CPU load presented by the uptime command or other monitoring tools. This CPU load is “the number of processes actively using the CPU or waiting for the CPU” so a single process running one core 100% will result in a load of approximately 1, a process using 100% of two cores 2, 4 processes each using one core 100% will add 4, and so on. If the load average over one minute is greater than the number of cores you have then things are waiting in the queue (or have been over that minute), which is fine for the occasional peak but not good if more constant. Of course you can have many more processes than cores because most of the time server processes are idle, waiting for incoming requests.

    It is only that simple for purely CPU-bottlenecked tasks though: the CPU load figure counts processes that are in the IO-Wait state: where they could be using CPU time in that jiffie¹ but were not able to because they were blocked waiting for IO operations to complete. This means that you can have a very high figure for load with little CPU activity. The rationale for measuring load this way is that as soon as the IO subsystem is unblocked a bit all those processes could immediately to want to eat some CPU resource. The CPU time isn't necessarily wasted, the IO-Wait state means the OS can let some other process use that process' allocated time, if there are other processes that are ready and not also blocked by IO.

    High load could also mean you are low on memory. If memory is full and a lot of swap space is used then there will be much paging going on imposing IO load (and making processes wait for pages they need to be reloaded), and there will be little or no room for the cache/buffers which would normally reduce, or at least smooth out, other IO bottlenecks a bit.

    ¹ a jiffie is the smallest amount of time the OS scheduler can allocate, usually 0.001 seconds (1/1000th of a second) on modern Linux on x86/a64 but it can be other values, 0.01 used to be the default, you will find different values in some embedded environments, on different architectures, etc.

    Thanked by 1ChaoscripT
  • @ChaoscripT said:
    Here is results from "top -c" command in Terminal,
    top - 08:48:36 up 12 days, 20:24, 0 users, load average: 5.29, 7.93, 8.11
    Tasks: 2 total, 1 running, 1 sleeping, 0 stopped, 0 zombie
    %Cpu(s): 13.8 us, 4.7 sy, 1.5 ni, 78.8 id, 0.4 wa, 0.0 hi, 0.7 si, 0.1 st
    MiB Mem : 64150.5 total, 7317.2 free, 17062.4 used, 39770.9 buff/cache
    MiB Swap: 4096.0 total, 2632.0 free, 1464.0 used. 31502.6 avail Mem

    With your cores mostly idle (not in active or IO-wait states) but load average high suggests all is quiet now but it wasn't a short while ago. The one-minute average being lower than the other two (5 minute, 15 minute) suggests that the peak load was more than five minutes ago.

    How high 5.29 or the other figures are, depends on how many CPU cores are actually available. Just one then those numbers are very high, if you have a 16 cores they are not a concern at all really (unless due to IO contention more than CPU demand).

    Only two tasks running seems odd for a hosting server. Do you have top running in a container so it can only see itself and your shell?!

    Thanked by 1ChaoscripT
  • Only two tasks running seems odd for a hosting server. Do you have top running in a container so it can only see itself and your shell?!

    What do you mean?

    With YABS Test I get this system information,

    Basic System Information:

    Processor : AMD EPYC 7502 32-Core Processor
    CPU cores : 12 @ 2495.312 MHz
    AES-NI : ✔ Enabled
    VM-x/AMD-V : ✔ Enabled
    RAM : 62.6 GiB
    Swap : 1024.0 MiB
    Disk : 2.5 TiB

    Regards.

  • vingohostvingohost Member, Host Rep

    Without complicated answers, The server will be running slow when the load average exceed the number of cpu cores assigned to the server, As it will start to queue processes and wait for the old processes to finish before processing the new ones.

    Thanked by 1yoursunny
  • @ChaoscripT said:

    Only two tasks running seems odd for a hosting server. Do you have top running in a container so it can only see itself and your shell?!

    What do you mean?

    With YABS Test I get this system information,

    Basic System Information:

    Processor : AMD EPYC 7502 32-Core Processor
    CPU cores : 12 @ 2495.312 MHz
    AES-NI : ✔ Enabled
    VM-x/AMD-V : ✔ Enabled
    RAM : 62.6 GiB
    Swap : 1024.0 MiB
    Disk : 2.5 TiB

    Regards.

    shared hosting?

  • @awooooool said:

    @ChaoscripT said:

    Only two tasks running seems odd for a hosting server. Do you have top running in a container so it can only see itself and your shell?!

    What do you mean?

    With YABS Test I get this system information,

    Basic System Information:

    Processor : AMD EPYC 7502 32-Core Processor
    CPU cores : 12 @ 2495.312 MHz
    AES-NI : ✔ Enabled
    VM-x/AMD-V : ✔ Enabled
    RAM : 62.6 GiB
    Swap : 1024.0 MiB
    Disk : 2.5 TiB

    Regards.

    shared hosting?

    Yes, Reseller hosting.

    Regards

Sign In or Register to comment.