Linux cpu utilisation and cpu load

K

kingsolly

New Member
Joined
Feb 2, 2013
Messages
1
Reaction score
0
Ok for all the people reading, lesson learnt don't ask a question from my phone this has now been updated.

I have a server that is on X86(intel) architecture
The server hardware is as follows
40 CPU's
528Gb Ram
7 Tb HDD
Running redhat

My Linux administrator informs me that the box is on the verge of falling over due to the overload of processing on it, The issue I have is that the monitoring tool I am using shows that CPU utilisation is only at 20% however the CPU load reflects that I am at 500% , for me this doesn't make sense as my understanding is that load reflects the physical load on the CPU's , shouldn't this then mean that CPU utilisation should actually reflect 100%.
the other interesting thing is that memory utilisation has never exceeded more than 32gb ram.
 
Last edited:
Are you sure that is 500GB of ram and 40 Cpus of what?

For me to believe this or take you seriously please state the servers details in full including the motherboard etc.

Also what are you running on it, the whole Denel ?
 
Also cost that's well above 200K in fact no that's above 300K depending on the Cpus and type of ram and mobo
 
Do you mean a CPU load of 500.0? CPU load is not measured in %.
 
CPU's or Core's ?

What do you get when you uptime or top ?

or paste the output of "cat /proc/cpuinfo" here
 
Last edited:
Now it is possible for the CPU to be 'busy' while not utilised. This means that it's not actually executing a command but can't proceed to the next command yet.

The most common reason for this is if it's waiting for data. This is called iowait and can be listed by the "top" command on Linux. This shows the % of time that the CPU is sitting waiting and not being used.

Most common reason I've seen for this is hard drives that are to slow for the work load on your server. This is normally a sign that it's time to throw out the hard drives and replace them all with SSDs.
 
or you're saturating your SATA/SAS bus with i/o, would this also cause high i/o wait @ambo?
 
Maybe it's 5 octacore CPUs, totalling 40 cores. Because if each core averages 62.5% util, but the statement is calculated using physical CPU, it could reflect 500% due to the 25 load on 40 "cores".
 
kingsolly said:
Ok for all the people reading, lesson learnt don't ask a question from my phone this has now been updated.

I have a server that is on X86(intel) architecture
The server hardware is as follows
40 CPU's
528Gb Ram
7 Tb HDD
Running redhat

My Linux administrator informs me that the box is on the verge of falling over due to the overload of processing on it, The issue I have is that the monitoring tool I am using shows that CPU utilisation is only at 20% however the CPU load reflects that I am at 500% , for me this doesn't make sense as my understanding is that load reflects the physical load on the CPU's , shouldn't this then mean that CPU utilisation should actually reflect 100%.
the other interesting thing is that memory utilisation has never exceeded more than 32gb ram.
Click to expand...

Wtf are you smoking, lad?
 
I'm sitting here and thinking what I am going to reply when he comes back with a screenshot, a photo, a invoice and a full lscpu, lshw and inxi -Fx report showing us 40 physical CPUs and 528Gb Ram.
 
Thor187 said:
I'm sitting here and thinking what I am going to reply when he comes back with a screenshot, a photo, a invoice and a full lscpu, lshw and inxi -Fx report showing us 40 physical CPUs and 528Gb Ram.
Click to expand...

Yeah something like Dafaq!
 
realistically i am sure we all know this guy is likely running a quad cpu board with 4x 10 core cpu's.. not sure why or how he got 528gb of ram.. 512Gb with similiar sized dimms yes, and then a single 16gb dimm.. strange..

anyhoo, anything this powerful should be running better performing storage than simple hdd's.. and also, get a new linux sysadmin if you do not trust him when he says the server is falling over.. as someone mentioned previously, the cpu's are more than likely waiting on the disks to their thing before the cpu's can do theirs, causing the high load..
 
I guess no one here is used to big corporate environments?

This is memory and CPU allocation of 1 of 11 servers running at our primary site.

Capture2.PNG
Capture.PNG

He could very well be running a VMWare environment which would explain X86.

Your Admin is probably talking about process load wait, which is not the same as CPU utilization. You probably have one process hogging CPU. Bottlenecks could be anywhere from disk to network connectivity.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X