Monitoring Memory and Processor

This chapter will introduce some of the basic commands used to investigate the configuration of a Linux server.  We will also look into some commands used to evaluate the status of the system.

It may also be necessary to determine the effects of adding new hardware or software to the server. Some examples of questions that might arise are:

* What are the top resource consumers on the system?

* Is the system I/O bound or CPU bound?

* How did adding disk and distributing the data affect I/O throughput?

* Did adding memory to the system improve memory performance?

First let's look at how to view the processor information in a Linux box.

CPU Related Information

Linux keeps information regarding the processor(s) in a server in a file called /proc/cpuinfo. The following are sample contents of the /proc/cpuinfo file from a basic Linux system:

# cat /proc/cpuinfo

The important information here is the vendor, model, CPU speed in MHz and cache size of this processor.  These are the factors most likely to affect performance.  Other information, like whether the processor is susceptible to some known CPU-related bugs, and whether it has a floating point unit (fpu) may also be useful, but have less effect on overall performance. The bogomips reference is an ambiguous speed rating calculated during boot-up. If the server contained multiple processors, the information in this listing would be repeated for each processor.

Display the Number of Processors in the Server

If we want to know how many processors a system has we can just look to see how many individual entries there are in the /proc/cpuinfo file.  If we wanted an easier way to check the number of processes we can use grep and wc to check the file for us like this:

# cat /proc/cpuinfo | grep processor | wc ?l
1

Here we have output the /proc/cpuinfo file with the cat command then used grep to eliminate all lines from teh output which do not contain the string 'processor'.  We then use wc ?l to count the number of lines left in the output.  The count is 1 so we know this system only has one processor.

Displaying the Total RAM on the Linux system

Linux stores memory related information in a file called /proc/meminfo. The meminfo file can be listed to see the current state of system memory:

# cat /proc/meminfo

MemTotal:       257124 kB
MemFree:         67388 kB
Buffers:         20516 kB
Cached:         124140 kB
SwapCached:          0 kB
Active:          51736 kB
Inactive:       108328 kB
HighTotal:           0 kB
HighFree:            0 kB
LowTotal:       257124 kB
LowFree:         67388 kB
SwapTotal:      524152 kB
SwapFree:       524152 kB
Dirty:               0 kB
Writeback:           0 kB
Mapped:          27752 kB
Slab:            26424 kB
Committed_AS:    64300 kB
PageTables:       1044 kB
VmallocTotal:  3874808 kB
VmallocUsed:      1260 kB
VmallocChunk:  3873360 kB
HugePages_Total:     0
HugePages_Free:      0

Some of the key points to this output are the MemTotal which is the total amount of memory in the system and the MemFree which shows how much unused memory is available.  The SwapTotal and SwapFree lines represent the same information but for swap (disk currently being used as memory.)

The Linux free command extracts and formats pertinent information from the meminfo.  While not as detailed free displays some big-picture information about the system:

# free -k

             total       used       free     shared    buffers     cached
Mem:        257124     189736      67388          0      20572     124140
-/+ buffers/cache:      45024     212100
Swap:       524152          0     524152

Some information in the above display is extracted directly from the meminfo file while other information is calculated. The display shows how much memory is being used in kilobytes, because the -k option has been specified.  If you prefer to see the information in megabytes you can use the ?m option.

Top Memory and CPU Users

Previously, the process status (ps) command was introduced. When the -u option is specified with the ps command, columns representing the percentage of CPU and the percentage of memory that a process is using are displayed as represented below by columns  %cpu and %mem.

# ps u

USER       PID %CPU %MEM   VSZ  RSS TTY      STAT START   TIME COMMAND
terry     4968  0.0  0.5  5644 1320 pts/6    S    11:21   0:00 -bash
terry     5120  0.0  0.2  3440  756 pts/6    R    11:42   0:00 ps u

The aux options can be added to the ps command to include all processes, not just the user?s own and build a complex command to display the commands using the most memory or CPU. 

An example of each complex command is included below:

Display Top CPU User:

# ps -aux | sort ?n +2 | tail -1

root      2201  0.2  3.2 36564 8464 ?        S    Oct17   2:18 /usr/X11R6/bin/X :0 -audit 0 -auth /var/gdm/:0.Xauth -nolisten tcp vt7

Display Top Memory User:

# ps -aux | sort -n +3 | tail -1

gdm       3718  0.0  3.5 20744 9012 ?        S    00:09   0:36 /usr/bin/gdmgreeter