4.3. Managing processes

4.3.5.1. Priority

The priority or importance of a job is defined by it's nice number. A program with a high nice number is friendly to other programs, other users and the system; it is not an important job. The lower the nice number, the more important a job is and the more resources it will take without sharing them.

Making a job nicer by increasing its nice number is only useful for processes that use a lot of CPU time (compilers, math applications and the like). Processes that always use a lot of I/O time are automatically rewarded by the system and given a higher priority (a lower nice number), for example keyboard input always gets highest priority on a system.

Defining the priority of a program is done with the nice command.

Most systems also provide the BSD renice command, which allows you to change the niceness of a running command. Again, read the man page for your system-specific information.

	Interactive programs
	It is NOT a good idea to nice or renice an interactive program or a job running in the foreground.

Use of these commands is usually a task for the system administrator. Read the man page for more info on extra functionality available to the system administrator.

4.3.5.2. CPU resources

On every Linux system, many programs want to use the CPU(s) at the same time, even if you are the only user on the system. Every program needs a certain amount of cycles on the CPU to run. There may be times when there are not enough cycles because the CPU is too busy. The uptime command is wildly inaccurate (it only displays averages, you have to know what is normal), but far from being useless. There are some actions you can undertake if you think your CPU is to blame for the unresponsiveness of your system:

• Run heavy programs when the load is low. This may be the case on your system during the night. See next section for scheduling.

• Prevent the system from doing unnecessary work: stop daemons and programs that you don't use, use locate instead of a heavy find, ...

• Run big jobs with a low priority

If none of these solutions are an option in your particular situation, you may want to upgrade your CPU. On a UNIX machine this is a job for the system admin.

4.3.5.3. Memory resources

When the currently running processes expect more memory than the system has physically available, a Linux system will not crash; it will start paging, or swapping, meaning the process uses the memory on disk or in swap space, moving contents of the physical memory (pieces of running programs or entire programs in the case of swapping) to disk, thus reclaiming the physical memory to handle more processes. This slows the system down enormously since access to disk is much slower than access to memory. The top command can be used to display memory and swap use. Systems using glibc offer the memusage and memusagestat commands to visualize memory usage.

If you find that a lot of memory and swap space are being used, you can try:

• Killing, stopping or renicing those programs that use a big chunk of memory

• Adding more memory (and in some cases more swap space) to the system.

• Tuning system performance, which is beyond the scope of this document. See the reading list in Appendix A for more.

4.3.5.4. I/O resources

While I/O limitations are a major cause of stress for system admins, the Linux system offers rather poor utilities to measure I/O performance. The ps, vmstat and top tools give some indication about how many programs are waiting for I/O; netstat displays network interface statistics, but there are virtually no tools available to measure the I/O response to system load, and the iostat command gives a brief overview of general I/O usage. Various graphical front-ends exist to put the output of these commands in a humanly understandable form.

Each device has its own problems, but the bandwidth available to network interfaces and the bandwidth available to disks are the two primary causes of bottlenecks in I/O performance.

Network I/O problems:

• Network overload:

  The amount of data transported over the network is larger than the network's capacity, resulting in slow execution of every network related task for all users. They can be solved by cleaning up the network (which mainly involves disabling protocols and services that you don't need) or by reconfiguring the network (for example use of subnets, replacing hubs with switches, upgrading interfaces and equipment).

• Network integrity problems:

  Occurs when data is transferred incorrectly. Solving this kind of problem can only be done by isolating the faulty element and replacing it.

Disk I/O problems:

• per-process transfer rate too low:

  Read or write speed for a single process is not sufficient.

• aggregate transfer rate too low:

  The maximum total bandwidth that the system can provide to all programs that run is not enough.

This kind of problem is more difficult to detect, and usually takes extra hardware in order to re-divide data streams over buses, controllers and disks, if overloaded hardware is cause of the problem. One solution to solve this is a RAID array configuration optimized for input and output actions. This way, you get to keep the same hardware. An upgrade to faster buses, controlers and disks is usually the other option.

If overload is not the cause, maybe your hardware is gradually failing, or not well connected to the system. Check contacts, connectors and plugs to start with.

4.3.5.5. Users

Users can be divided in several classes, depending on their behavior with resource usage:

• Users who run a (large) number of small jobs: you, the beginning Linux user, for instance.

• Users who run relatively few but large jobs: users running simulations, calculations, emulators or other programs that eat a lot of memory, and usually these users have accompanying large data files.

• Users who run few jobs but use a lot of CPU time (developers and the like).

You can see that system requirements may vary for each class of users, and that it can be hard to satisfy everyone. If you are on a multi-user system, it is useful (and fun) to find out habits of other users and the system, in order to get the most out of it for your specific purposes.

4.3.5.6. Graphical tools

For the graphical environment, there are a whole bunch of monitoring tools available. Below is a screen shot of the Gnome System Monitor, which has features for displaying and searching process information, and monitoring system resources:

There are also a couple of handy icons you can install in the task bar, such as a disk, memory and load monitor. xload is another small X application for monitoring system load. Find your favorite!

4.3.5.7. Interrupting your processes

As a non-privileged user, you can only influence your own processes. We already saw how you can display processes and filter out processes that belong to a particular user, and what possible restrictions can occur. When you see that one of your processes is eating too much of the system's resources, there are two things that you can do:

1. Make the process use less resources without interrupting it;

2. Stop the process altogether.

In the case that you want the process to continue to run, but you also want to give the other processes on the system a chance, you can renice the process. Appart from using the nice or renice commands, top is an easy way of spotting the troublesome process(es) and reducing priority.

Identify the process in the "NI" column, it will most likely have a negative priority. Type r and enter the process ID of the process that you want to renice. Then enter the nice value, for instance "20". That means that from now on, this process will take 1/5 of the CPU cycles at the most.

Examples of processes that you want to keep on running are emulators, virtual machines, compilers and so on.

If you want to stop a process because it hangs or is going totally berserk in the way of I/O consumption, file creation or use of other system resources, use the kill command. If you have the opportunity, first try to kill the process softly, sending it the SIGTERM signal. This is an instruction to terminate whatever it is doing, according to procedures as described in the code of the program:

joe:~> ps -ef | grep mozilla
joe    25822	1  0 Mar11 ?	00:34:04 /usr/lib/mozilla-1.4.1/mozilla-

joe:~> kill -15 25822

In the example above, user joe stopped his Mozilla browser because it hung.

Some processes are a little bit harder to get rid of. If you have the time, you might want to send them the SIGINT signal to interrupt them. If that does not do the trick either, use the strongest signal, SIGKILL. In the example below, joe stops a Mozilla that is frozen:

joe:~> ps -ef | grep mozilla
joe    25915	1  0 Mar11 ?	00:15:06 /usr/lib/mozilla-1.4.1/mozilla-

joe:~> kill -9 25915

joe:~> ps -ef | grep 25915
joe	2634 32273 0 18:09 pts/4   00:00:00 grep 25915

In such cases, you might want to check that the process is really dead, using the grep filter again on the PID. If this only returns the grep process, you can be sure that you succeeded in stopping the process.

Among processes that are hard to kill is your shell. And that is a good thing: if they would be easy to kill, you woud loose your shell every time you type Ctrl-C on the command line accidentally, since this is equivalent to sending a SIGINT.

	UNIX without pipes is almost unthinkable
	The usage of pipes (|) for using output of one command as input of another is explained in the next chapter, Chapter 5.

In a graphical environment, the xkill program is very easy to use. Just type the name of the command, followed by an Enter and select the window of the application that you want to stop. It is rather dangerous because it sends a SIGKILL by default, so only use it when an application hangs.