Chapter 14 - Operating System Information
This chapter describes statistics on kernel (server operating system) activity, and describes configuration information about internal operating system parameters on VINES servers. If you are familiar with standard UNIX system activity statistics, you will easily recognize the statistics and information described in this chapter.
Viewing Operating System Statistics and Information
To view operating system statistics and configuration information, choose SHOW OS information from the VINES Network Summary menu. The OS Information screen appears as follows.
The screen shows a table, each row of which shows the statistics or configuration information within a specific category. The top of this screen displays the server's name, its network ID, its platform type, and the number of CPUs. Table 14-1 describes some common platform types.
Servers with an SMP platform type have more than one CPU. For these servers, the statistics in the following rows are calculated based on the average for all the CPUs:
CPU Utilization Swaps Paging/Switches
To return to the VINES Network Summary menu, press ESC or F3.
The statistics that appear on the screen show you operating system activity over the last five minutes and are updated every minute. For example, the user statistic at 10:00 shows the percentage of time that the CPU spent servicing requests from processes that run under UNIX since 9:55. A minute later, this statistic is calculated again and shows activity since 9:56.
All of the averages in this chapter are flat averages.
The sections that follow explain the data on the OS Information screen.
This section describes CPU utilization statistics, which tell you how much time the CPU spent processing the following events:
Requests from services and operating system routines I/O requests
user
The percentage of time that the CPU spent running processes that run in user space under UNIX. These processes include services and third-party applications.
sys
The percentage of time that the CPU spent running in system mode. When the CPU runs in system mode, it services requests from kernel routines. Examples of kernel routines include the routines that implement communication protocols, such as VINES, TCP/IP, and AppleTalk protocols; routines that handle the Server Message Block (SMB) requests from DOS, Windows and OS/2 workstations; and operating system routines.
wio
The percentage of time that the CPU was idle, but at least one process was waiting for an I/O operation, such as a disk read or write, to complete. A significant percentage, such as 20 percent or greater, indicates that a disk bottleneck exists. In this case, correlate this statistic with the AvgWait disk usage statistic, which is described in Chapter 9. A high percentage may indicate that too many services were added to a disk.
idle
The percentage of time that the CPU was completely idle (no processes were waiting to run or waiting for I/O operations to complete).
This section describes run/swap queue utilization statistics which tell you the average number of processes that occupied the run and swap queues over the last five minutes and the percentage of time that the queues were occupied over the last five minutes.
The run queue is an internal list of processes that need to be serviced by the CPU.
The swap queue is an internal list of processes that are swapped from main memory to the swap area of the disk. See Chapter 3 for an explanation of swapping.
If the swap queue is occupied, the statistics described in this section will be non-zero. This indicates that the server is swapping. Swapping can cause performance degradation on a server and should be avoided. Chapter 3 explains how to deal with swapping problems.
runq
The average number of processes in the run queue while the queue was occupied. This statistic reflects the load on the processor.
runocc
The percentage of time that processes occupied the run queue over a 5-minute period. This statistic reflects the load on the processor.
swpq
The average number of processes in the swap queue while it was occupied. If this statistic is non-zero, swapping is taking place. Correlate this statistic with the Swavg statistic, which is described in Chapter 3.
swpocc
The percentage of time that processes occupied the swap queue over a 5-minute period.
This section describes UNIX system buffer utilization statistics. UNIX system buffers differ from the cache buffers in VINES file system cache, which you can view through the SHOW file system statistics function. VINES file system cache is used to store large- or medium-sized files for fast access by the VINES file system or VINES communications software. UNIX system buffers are reserved for caching small files, directories, and other files that cannot be stored in VINES file system cache.
When interpreting the statistics in this section, keep in mind that servers that have fast CPUs (such as an Intel 486 processor) and disk drives are better able to process heavy buffer demands.
Only VINES file system cache is configurable; UNIX system buffers are not. If the statistics described in this section indicate heavy levels of activity, consider moving heavily used file services or database programs off the server, or reducing the number of users of the file services or database programs.
See Chapter 8 for more information on VINES file system statistics.
brd/s
The average number of times per second that data was read from disk into UNIX system buffers. Values less than 10 indicate a light amount of read activity. Values between 10 and 20 indicate a moderate amount of read activity. Values above 20 indicate a heavy amount of read activity.
rdcach
The percentage of read requests for UNIX system buffers that the kernel was able to service using existing UNIX system buffers, without reading new data from disk. For most servers, 85 percent or higher is good. If you need extremely fast access to data, your goal should be 95 percent or higher.
bwr/s
The average number of times per second that data was written from UNIX system buffers to disk. Values less than 10 indicate a light amount of write activity. Values between 10 and 20 indicate a moderate amount of write activity. Values above 20 indicate a heavy amount of write activity.
wrcach
The percentage of write requests for UNIX system buffers that the kernel was able to service using existing UNIX system buffers, without writing new data to disk.
This section describes statistics that apply to swapping activity. See Chapter 3 for a description of swapping.
The average number of 4 KB pages that were swapped or paged from disk to main memory per second over the last five minutes. Values less than 10 indicate a light amount of activity. Values between 10 and 20 indicate a moderate amount of activity. Values above 20 indicate a heavy amount of activity.
Correlate this statistic with the bswpo/s statistic and the Swavg statistic described in Chapter 3. If all three of these statistics indicate a heavy amount of activity, consider ways to reduce paging and swapping activity on the server. Chapter 3 describes ways to reduce this kind of activity.
bswpo/s
The average number of 4 KB pages that were swapped or paged from main memory to disk per second over the last five minutes. Values less than 10 indicate a light amount of activity. Values between 10 and 20 indicate a moderate amount of activity. Values above 20 indicate a heavy amount of activity.
Correlate this statistic with the bswpi/s statistic and the Swavg statistic described in Chapter 3. If all three of these statistics indicate a heavy amount of activity, consider ways to reduce paging and swapping activity on the server. Chapter 3 describes ways to reduce this kind of activity.
The amount of available swap space, in 512-byte blocks, on disk, and the total amount of swap space on disk. This statistic appears thus:
available/total
where available is the available swap space and total is the total amount of swap space. For example, if freeswap displays 8000/9000, then 8000 blocks of swap space are available out of a total of 9000 blocks. One thousand blocks are in use.
If the server attempts to page or swap a process, such as a service, and the current swap space is insufficient, the process crashes. See Chapter 3 for more information on the Swavg statistic. See Chapter 15 for more information on increasing swap space.
This section describes statistics that apply to paging activity, process switching activity, and system calls.
See Chapter 3 for a description of paging.
The kernel performs process switching when one UNIX process, such as a VINES service, is in sleeping state and another process needs to be executed. A process is in sleeping state when it cannot keep executing, such as when it awaits the completion of an I/O event.
UNIX processes can issue system calls to perform a wide variety of operations, ranging from I/O operations to main memory allocation.
vflts/s
The average number of validity faults per second over the last five minutes. A validity fault occurs when a process attempts to access a page that was part of the virtual address space of the process but did not have a physical page assigned to it. Validity faults are not errors. They occur as part of the normal operation of UNIX.
This statistic increases noticeably when services start on the server, when services are swapping or paging, when services are heavily used, or as the number of services on the server increase. If you notice a brief increase in this statistic just after services are started on the server, do not be concerned. However, if you notice sharp and consistent increases in this statistic at any other time, correlate this statistic with the bswpi/s and bswpo/s statistics described earlier in this chapter, and the Swavg statistic described in Chapter 3. If all three of these statistics indicate a heavy amount of activity, consider ways to reduce paging and swapping activity on the server. Chapter 3 describes ways to reduce this kind of activity.
swtch/s
The average number of process switches per second over the last five minutes. A value greater than 60 indicates that your server is heavily loaded. In this case, consider moving heavily used services to other servers, stopping unused services, reducing the number of users of services on the server, or adding more RAM.
Note: The amount of process switching load that a server handles depends on its CPU. The faster the CPU, the more process switching load the server can handle.
syscal/s
The average number of UNIX system calls that processes issued per second over the last five minutes. As the amount of service activity on the server increases, this value increases.
This section describes information about the contents of internal UNIX configuration tables that are displayed on the OS Information screen. Information is displayed in the following format:
current/maximum
where current is the current number of entries in the table and maximum is the maximum number of entries that can be in the table. Note that the current number of entries changes.
Process
The number of process table entries that the server operating system allocates. Each active process uses one entry in the process table. There can be up to 200 processes running at once on the server. Most VINES and third-party services run as one process, but some services, such as the VINES Network Mail service, run as more than one process.
Inode
The number of inode table entries that the operating system allocates. Each open file, current directory for a user, or mounted file system requires an inode entry from the inode table. A diskette is a mounted file system. The default maximum size of this table is 400 entries; however, if this limit is reached or exceeded the server can allocate more entries. Thus, you may notice a maximum limit that is greater than 400 on heavily-used servers.
File
The number of open file table entries that the operating system allocates. Each open file has an entry in the open file table. There can be up to 700 entries in this table.
If UNIX runs out of open file table entries, a message indicating file table overflow appears at the server console.
Note that a file table overflow does not result in a server crash or a loss of data. However, an overflow prevents files being opened until at least one open file is closed.
This section describes statistics on available RAM (main memory) and swap space on disk. Swap space is used for both paging and swapping.
realmem
The total amount of RAM, in kilobytes, configured in the server.
The amount of executable space, in kilobytes, that is available for VINES services to run in, and the total amount of executable space in the server. Executable space is the part of RAM in which UNIX processes, such as VINES services, run. See Chapter 3 for a description of executable space.
This statistic appears in the following format:
available/total
where available is the available executable space and total is the total amount of executable space. For example, if freemem displays 176/8802, 176 KB of executable space is available out of a total of 8802 KB. 8626 KB of executable space is in use by VINES services.
Correlate this statistic with the Swavg statistic, which tells you whether the server is swapping, and the Siz statistic, which tells you how much executable space an individual service is using. See Chapter 3 for a description of the Swavg statistic. See Chapter 7 for a description of the Siz statistic.
The freemem statistic may show a small amount of available executable space even though Swavg indicates that the server is swapping. The server operating system reserves some executable space to perform overhead functions.