 |
DECamds User's Guide
3.5 Node Summary Window
The Node Summary window shown in Figure 3-6 displays a high-level
graphic summary of node resource demands on the CPU, memory, and I/O.
Figure 3-6 Node Summary Window
To open a Node Summary window, do one of the following:
- In the System Overview window, double-click on any node name. You
can also click MB3 on any node name, and choose Display from the menu.
- In the Event Log window, double-click on any node name. You can
also click MB3 on an event that is related to node summary data, and
choose Display from the menu.
Dynamic bar graphs display the current values for each field. Peak
values are also displayed from when DECamds begins collecting node
summary data. A peak value is typically the highest value received;
however, for the Free Memory field it is the lowest value received.
You can open the following windows from the Node Summary Window by
double-clicking in the space for each category:
CPU Summary
CPU Modes Summary
Memory Summary
I/O summary
Table 3-6 describes the Node Summary window data fields.
Table 3-6 Node Summary Window Data Fields
| Field |
Displays |
|
Hardware Model
|
The system hardware model name.
|
|
Operating System
|
The name and version of the operating system.
|
|
Uptime
|
The time since last reboot measured in days, hours, minutes, and
seconds.
|
|
Memory
|
The total amount of physical memory found on the system.
|
|
CPUs
|
The number of active CPUs on the node.
|
|
CPU Process State Queues
|
One of the following:
|
COM
|
Sum of the queue lengths of processes in the COM and COMO states.
|
|
WAIT
|
Sum of the queue lengths of processes in the MWAIT, COLPG, CEF, PFW,
and FPG states.
|
|
|
CPU Modes
|
The CPU usage by mode (kernel, executive, supervisor, user, interrupt,
compatibility, multiprocessor synchronization, and null). On symmetric
multiprocessing (SMP) nodes, percentages are averaged across all the
CPUs and displayed as one value.
|
|
Page Faults
|
The rate of system hard and soft page faulting, as well as peak values
seen during a DECamds session. System page faults are taken from kernel
processes.
|
|
Memory
|
The histogram listing memory distribution (Free, Used, Modified, Bad)
as absolute values of number of thousands of pages or pagelets. Peak
values are also listed, with Free using lowest seen value as peak.
|
|
I/O
|
The histogram listing Buffer, Direct, and Page Write I/O rates per
second. Also included is the peak value seen.
|
DECamds detects the following node events and displays them in the
Event Log window. Node is replaced by the name of the node to
which the event is related.
HIBIOR, node buffered I/O rate is high
HICOMQ, node many processes waiting for CPU
HIDIOR, node direct I/O rate is high
HIHRDP, node hard page fault rate is high
HIMWTQ, node process waiting in MWAIT
HINTER, node interrupt mode time is high
HIPWIO, node paging write I/O rate is high
HIPWTQ, node many processes waiting in COLPG, PFW, or FPG
HITTLP, node total page fault rate is high
HMPSYN, node MP synchronization mode time is high
HISYSP, node system page fault rate is high
LOMEMY, node free memory is low
NOPROC, node cannot find process names process
3.6 Process I/O Summary Window
The Process I/O Summary window shown in Figure 3-7 displays summary
statistics about process I/O rates and quotas. Use the Process I/O
Summary window to display information about I/O issues that might be
caused by I/O-intensive programs or I/O bottlenecks.
Note
DECamds does not yet support kernel threads. If you use threaded
processes, DECamds displays only the top thread.
|
Figure 3-7 Process I/O Summary Window
To open a Process I/O Summary window, do one of the following:
- In the Node Summary window, double-click in the I/O area.
- On the View menu in the Single Disk Summary window, choose Display
Process I/O Summary.
- In the System Overview window, double-click on the BIO or DIO
fields for any node. You can also click MB3 on any field for any node,
choose Display from the menu, and choose Process I/O Summary from the
submenu.
- To open a Process I/O Summary window for every node in a group, in
the System Overview window, click MB3 on a group line, choose Display
from the menu, and choose Process I/O Summary from the submenu.
- In the Event Log window, click MB3 on any process I/O-related
event, and choose Display from the menu.
You can open a window about a specific process in the Process I/O
Summary window by double-clicking on the process name.
Table 3-7 describes the Process I/O Summary window data fields.
Table 3-7 Process I/O Summary Window Data Fields
| Field |
Displays |
|
PID
|
The process identifier, a 32-bit value that uniquely identifies a
process.
|
|
Process Name
|
The current process name.
|
|
Direct I/O Rate (DIO)
|
The rate at which I/O transfers occur between the system devices and
the pages or pagelets that contain the process buffer that the system
locks in physical memory.
|
|
Buffered I/O Rate (BIO)
|
The rate at which I/O transfers occur between the process buffer and an
intermediate buffer from the system buffer pool.
|
|
Paging I/O Rate (PIO)
|
The rate of read attempts necessary to satisfy page faults (also known
as Page Read I/O or the Hard Fault Rate).
|
|
Open Files
|
The number of open files.
|
|
Direct I/O Limit Remaining (DIO)
|
The number of remaining direct I/O limit operations available before
the process reaches its quota. DIOLM quota is the maximum number of
direct I/O operations a process may have outstanding at one time.
|
|
Buffered I/O Limit Remaining (BIO)
|
The number of remaining buffered I/O operations available before the
process reaches its quota. BIOLM quota is the maximum number of
buffered I/O operations a process may have outstanding at one time.
|
|
Byte Limit Remaining (Bytes)
|
The number of buffered I/O bytes available before the process reaches
its quota. BYTLM is the maximum number of bytes of nonpaged system
dynamic memory that a process can claim at one time.
|
|
Open File Limit Remaining (Files)
|
The number of additional files the process can open before reaching its
quota. FILLM quota is the maximum number of files that can be opened
simultaneously by the process, including active network logical links.
|
DECamds detects the following process I/O-related events and displays
them in the Event Log window. Node is replaced by the name of
the node to which the event is related. Process is replaced by
the name of the process to which the event is related.
LOBIOQ, node process has used most of its BIOLM process quota
LOBYTQ, node process has used most of its BYTLM job quota
LODIOQ, node process has used most of its DIOLM process
quota
LOFILQ, node process has used most of its FILLM job quota
PRBIOR, node process buffered I/O rate is high
PRDIOR, node process direct I/O rate is high
PRPIOR, node process paging I/O rate is high
3.7 CPU Modes Summary Window
The CPU Modes Summary window shown in Figure 3-8 displays more
detailed summary statistics about CPU mode usage than the Node Summary
window. Use the CPU Modes Summary window to diagnose issues that may be
caused by CPU-intensive users or CPU bottlenecks.
Figure 3-8 CPU Modes Summary Window
To open a CPU Modes Summary window, do one of the following:
- In the Node Summary window, double-click MB1 in the CPU Modes area.
You can also click MB3, and choose Display from the menu.
- In the Node Summary window View menu, choose Display Modes Summary.
You can open a window about a specific process in the CPU Modes Summary
window by double-clicking on the process name.
Table 3-8 describes the CPU Modes Summary window data fields.
Table 3-8 CPU Modes Summary Window Data Fields
| Field |
Displays |
|
CPU ID
|
A decimal value representing the identity of a process in a
multiprocessing system. On a uniprocessor, this value will always be
CPU #00.
|
|
Capabilities
|
One of the the following CPU capabilities: Primary, Quorum, Run, or
Vector.
|
|
State
|
One of the following CPU states: Boot, Booted, Init, Rejected,
Reserved, Run, Stopped, Stopping, or Timeout.
|
|
Mode
|
One of the following values for CPU modes supported for the
architecture: Compatibility, Executive, Interrupt, Kernel, MP Synch,
Null, Supervisor, or User. Note: Compatibility mode does not exist on
OpenVMS Alpha systems.
|
|
% Used
|
A bar graph, by CPU, representing the percentage of the CPU utilization
for each mode.
|
|
PID
|
The process identifier value of the process that is using the CPU. If
the PID is unknown to the console application, the internal PID (IPID)
will be listed.
|
|
Name
|
The process name of the process found in the CPU. If no process is
found in the CPU, this will be listed as *** None ***.
|
|
Rate
|
A numerical percentage of CPU time for each mode.
|
|
Peak
|
The peak CPU usage determined for each mode.
|
3.8 CPU Summary Window
The CPU Summary window shown in Figure 3-9 displays summary
statistics about process CPU usage issues that might be caused by
CPU-intensive users or CPU bottlenecks.
Figure 3-9 CPU Summary Window
To open a CPU Summary window, do one of the following:
- In the System Overview window, double-click on the CPU field of any
node. You can also click MB3 on an event that is related to CPU usage,
choose Display from the menu, and choose CPU Summary from the list.
- In the Node Summary window, double-click on CPU Process State
Queues.
- In the Event Log window, click MB3 on an event that is related to
CPU usage, choose Display from the menu, and choose CPU Summary from
the list.
You can open a window about a specific process in the CPU Summary
window by double-clicking on the process name.
Table 3-9 describes the CPU Summary window data fields.
Table 3-9 CPU Summary Window Data Fields
| Field |
Displays |
|
PID
|
The process identifier, a 32-bit value that uniquely identifies a
process.
|
|
Name
|
The process name.
|
|
Priority
|
Computable
(xx) and base
(yy) process priority in the format
xx/yy.
|
|
State
|
One of the values listed under the Single Process Summary description
in Table 3-11.
|
|
Rate
|
The percent of CPU time used by this process. This is the ratio of CPU
time to elapsed time. The CPU rate is also displayed in the bar graph.
|
|
Wait
|
The percent of time the process is in the COM or COMO state.
|
|
Time
|
The amount of actual CPU time charged to the process.
|
DECamds detects the following CPU-related events and displays them in
the Event Log window. Node is replaced by the name of the node
to which the event is related. Process is replaced by the name
of the process to which the event is related.
PRCCOM, node process waiting in COM or COMO
PRCCVR, node process has high CPU rate
PRCMWT, node process waiting in MWAIT
PRCPWT, node process waiting in COLPG, PFW, or FPG
3.9 Memory Summary Window
The Memory Summary window shown in Figure 3-10 displays memory usage
data for processes on a node so that you can identify processes that
use large amounts of memory or have high page fault rates.
Figure 3-10 Memory Summary Window
To open a Memory Summary window, do one of the following:
- In the Node Summary window, double-click on the Page Faults or
Memory area. You can also click MB3 on the Page Faults or Memory area,
and choose Display from the menu.
- In the View menu of the Node Summary window, choose Display Memory
Summary.
- In the System Overview window, double-click on the Memory field for
any node. You can also click MB3 on any field for any node, choose
Display from the pop-up menu, and choose Memory Summary from the
submenu.
- To display a memory summary of every node in a group from the
System Overview window, click MB3 on the group line, choose Display
from the menu, and choose Memory Summary from the submenu.
- In the Event Log window, click MB3 on an event related to memory
usage, and choose Display from the menu.
You can open a window about a specific process in the Memory Summary
window by double-clicking on the process name.
Table 3-10 describes the Memory Summary window data fields.
Table 3-10 Memory Summary Window Data Fields
| Field |
Displays |
|
PID
|
The process identifier, a 32-bit value that uniquely identifies a
process.
|
|
Process Name
|
The process name.
|
|
Working Set Count
1
|
The number of physical pages or pagelets of memory that the process is
using. The bar graph represents the percentage of working set count
used to the working set extent.
|
|
Working Set Size
1
|
The number of pages or pagelets of memory the process is allowed to
use. This value is periodically adjusted by the operating system based
on analysis of page faults relative to CPU time used. When the value
increases in large units, this indicates a process is receiving a lot
of page faults and its memory allocation is increasing.
|
|
Working Set Extent
1
|
The number of pages or pagelets of memory in the process's WSEXTENT
quota as defined in the user authorization file (UAF). The number of
pages or pagelets will not exceed the value of the system parameter
WSMAX.
|
|
Page Fault Rate
|
The number of page faults per second for the process. The bar graph
represents a relative number of page faults per second.
|
|
Page Fault I/O
|
The rate of read attempts necessary to satisfy page faults (also known
as Page Read I/O or the Hard Fault Rate).
|
1Working Set Value = Total Physical Memory / Maximum Process
Count
DECamds detects the following memory-related events and displays them
in the Event Log window. Node is replaced by the name of the
node to which the event is related. Process is replaced by the
name of the process to which the event is related.
LOWEXT, node process working set extent is too small
LOWSQU, node process working set quota is too small
PRPGFL, node process high page fault rate
PRPIOR, node process paging I/O rate is high
3.10 Single Process Summary Window
The Single Process Summary window shown in Figure 3-11 displays
summary data about a process, including Execution Rates, Process Quotas
in Use, Wait States, and Job Quotas in Use.
Figure 3-11 Single Process Summary Window
To open a Single Process Summary window, do one of the following:
- In any window that displays processes (CPU, CPU Modes, Memory,
Process I/O, and Single Lock Summary), double-click on any field. You
can also click MB3 on any field in a process line, and choose Display
from the pop-up menu.
- You can also click on any field in a process line, and choose
Display Process from the View menu.
- In the Event Log window, double-click on a process-related event.
You can also click MB3 on a process-related event, choose Display from
the menu, and choose Single Process in the dialog box.
Table 3-11 describes the Single Process Summary window data fields.
Table 3-11 Single Process Summary Window Data Fields
| Field |
Displays |
|
Process name
|
The name of the process.
|
|
Username
|
The user name of the user owning the process.
|
|
Account
|
The string assigned to the user by the system manager.
|
|
UIC
|
The user identification code (UIC), a pair of numbers or character
strings designating the group and user.
|
|
PID
|
The process identifier, a 32-bit value that uniquely identifies a
process.
|
|
Owner ID
|
The PID of the process that created the process displayed in the
window. If 0, then the process is a parent process.
|
|
PC
|
The program counter. On OpenVMS VAX systems, this is the address of the
next instruction the CPU will execute. On OpenVMS Alpha systems, this
value is displayed as 0, because the data is not readily available to
the Data Provider node.
|
|
PSL
|
The processor status longword (PSL). On OpenVMS VAX systems, this
indicates the current processor mode (user, kernel, and so on) and its
interrupt level. On OpenVMS Alpha systems, this value is displayed as
0, because the data is not readily available to the Data Provider node.
|
|
Priority
|
The computable and base priority of the process. Priority is an integer
between 0 and 31. Processes with higher priority get more CPU time.
|
|
State
|
One of the following process states:
|
CEF
|
Common Event Flag, waiting for a Common Event Flag
|
|
COLPG
|
Collided Page Wait, involuntary wait state; likely indicates a memory
shortage, waiting for hard page faults
|
|
COM
|
Computable; ready to execute
|
|
COMO
|
Computable Outswapped, COM, but swapped out
|
|
CUR
|
Current, currently executing in a CPU
|
|
FPW
|
Free Page Wait, involuntary wait state; likely indicates a memory
shortage
|
|
LEF
|
Local Event Flag, waiting for a Local Event Flag
|
|
LEFO
|
Local Event Flag Outswapped; LEF, but outswapped
|
|
HIB
|
Hibernate, voluntary wait state requested by the process; it is inactive
|
|
HIBO
|
Hibernate Outswapped, hibernating but swapped out
|
|
MWAIT
|
Miscellaneous Resource Wait, involuntary wait state; possibly caused by
a shortage of a systemwide resource such as no page or swap file
capacity or synchronizations for single threaded code
|
|
PFW
|
Page Fault Wait, involuntary wait state; possibly indicates a memory
shortage, waiting for hard page faults
|
|
RWAST
|
Resource Wait State, waiting for delivery of an asynchronous system
trap (AST) that signals a resource availability; usually an I/O is
outstanding or a process quota is exhausted
|
|
RWBRK
|
Resource Wait for BROADCAST to finish
|
|
RWCAP
|
Resource Wait for CPU Capability
|
|
RWCLU
|
Resource Wait for Cluster Transition
|
|
RWCSV
|
Resource Wait for Cluster Server Process
|
|
RWIMG
|
Resource Wait for Image Activation Lock
|
|
RWLCK
|
Resource Wait for Lock ID data base
|
|
RWMBX
|
Resource Wait on MailBox, either waiting for data in mailbox (to read)
or waiting to place data (write) into a full mailbox (some other
process has not read from it; mailbox is full so this process cannot
write).
|
|
RWMPB
|
Resource Wait for Modified Page writer Busy
|
|
RWMPE
|
Resource Wait for Modified Page list Empty
|
|
RWNPG
|
Resource Wait for Non Paged Pool
|
|
RWPAG
|
Resource Wait for Paged Pool
|
|
RWPFF
|
Resource Wait for Page File Full
|
|
RWQUO
|
Resource Wait for Pooled Quota
|
|
RWSCS
|
Resource Wait for System Communication Services
|
|
RWSWP
|
Resource Wait for Swap File space
|
|
SUSP
|
Suspended, wait state process placed into suspension; it can be resumed
at the request of an external process
|
|
SUSPO
|
Suspended Outswapped, suspended but swapped out
|
|
|
WS global pages
|
The shared data or code between processes, listed in pages or pagelets.
|
|
WS private pages
|
The amount of accessible memory, listed in pages or pagelets.
|
|
WS total pages
|
The sum of global and private pages or pagelets.
|
|
WS size
|
The working set size, number of pages or pagelets of memory the process
is allowed to use. This value is periodically adjusted by the operating
system based on analysis of page faults relative to CPU time used. When
it increases in large units, this indicates a process is taking a lot
of page faults and its memory allocation is increasing.
|
|
WSdef
|
The working set default, the initial limit to the number of physical
pages or pagelets of memory the process can use. This parameter is
listed in the user authorization file (UAF); discrepancies between the
UAF value and the displayed value are due to page/longword boundary
rounding or other adjustments made by the operating system.
|
|
WSquo
|
The working set quota, the maximum amount of physical pages or pagelets
of memory the process can lock into its working set. This parameter is
listed in the UAF; discrepancies between the UAF value and the
displayed value are due to page/longword boundary rounding or other
adjustments made by the operating system.
|
|
WSextent
|
The working set extent, the maximum number of physical pages or
pagelets of memory the system will allocate for the process. The system
provides memory to a process beyond its quota only when it has an
excess of free pages and can be recalled if necessary. This parameter
is listed in the UAF; any discrepancies between the UAF value and the
displayed value are due to page/longword boundary rounding or other
adjustments made by the operating system.
|
|
Images activated
|
The number of times an image is activated.
|
|
Mutexes held
|
The number of mutual exclusions (mutexes) held. Persistent values other
than zero (0) require analysis. A mutex is similar to a lock but is
restricted to one CPU. When a process holds a mutex, its priority is
temporarily incremented to 16.
|
| Execution Rates |
|
CPU
|
The percent of CPU time used by this process. This is the ratio of CPU
time to elapsed time. CPU rate is also displayed in the bar graph.
|
|
Direct I/O
|
The rate at which I/O transfers take place from the pages or pagelets
containing the process buffer that the system locks in physical memory
to the system devices.
|
|
Buffered I/O
|
The rate at which I/O transfers take place for the process buffer from
an intermediate buffer from the system buffer pool.
|
|
Paging I/O
|
The rate of read attempts necessary to satisfy page faults. This is
also known as Page Read I/O or the Hard Fault Rate.
|
|
Page Faults
|
The page faults per second for the process. The bar graph visually
represents page faults per second.
|
|
|
|
| Process Quotas in Use1 |
|
DIOLM
|
Direct I/O Limit. A bar graph representing current count of DIOs used
with respect to the limit that can be attained.
|
|
BIOLM
|
Buffered I/O Limit. A bar graph representing current count of BIOs used
with respect to the limit that can be attained.
|
|
ASTLM
|
Asynchronous System Traps Limit. A bar graph representing current count
of ASTs used with respect to the limit that can be attained.
|
|
CPU
|
CPU Time Limit. A bar graph representing current count of CPU time used
with respect to the limit that can be attained. If the limit is 0, then
this value is not used.
|
| Wait States2 |
|
Compute
|
A relative value indicating that the process is waiting for CPU time.
The included states are COM, COMO, RWCAP.
|
|
Memory
|
A relative value indicating that the process is waiting for a page
fault that requires data to be read from disk; common during image
activation. The included states are PFW, COLPG, FPG, RWPAG, RWNPG,
RWMPE, RWMPB.
|
|
Direct I/O
|
A relative value indicating that the process is waiting for data to be
read from or written to a disk. The included state is DIO.
|
|
Buffered I/O
|
A relative value indicating that the process is waiting for data to be
read from or written to a slower device such as a terminal, line
printer, or mailbox. The included state is BIO.
|
|
Control
|
A relative value indicating that the process is waiting for another
process to release control of some resource. The included states are
CEF, MWAIT, LEF, LEFO, RWAST, RWMBX, RWSCS, RWCLU, RWCSV, RWUNK, and
LEF waiting for a ENQ.
|
|
Quotas
|
A relative value indicating that the process is waiting because the
process has exceeded some quota. The included states are QUOTA and
RWAST_QUOTA.
|
|
Explicit
|
A relative value indicating that the process is waiting because the
process asked to wait, such as a hibernate system service. The included
states are HIB, HIBO, SUSP, SUSPO, and LEF waiting for a TQE.
|
| Job Quotas in Use |
|
FILLM
|
File Limit. A bar graph representing current number of open files with
respect to the limit that can be attained.
|
|
PGFLQUO
|
Page File Quota. A bar graph representing current number of disk blocks
in page file that the process can use with respect to the limit that
can be attained.
|
|
ENQLM
|
Enqueue Limit. A bar graph representing current count of resources
(lock blocks) queued with respect to the limit that can be attained.
|
|
TQELM
|
Timer Queue Entry Limit. A bar graph representing current count of
timer requests with respect to the limit that can be attained.
|
|
PRCLM
|
Process Limit. A bar graph representing current count of subprocesses
created with respect to the limit that can be attained.
|
|
BYTLM
|
Buffered I/O Byte Limit. A bar graph representing current count of
bytes used for buffered I/O transfers with respect to the limit that
can be attained.
|
|
Image Name
|
The name of the currently executing image, if available. If this field
does not appear, then the data is not resident in memory.
|
1When you display the SWAPPER process, no values are listed
in this section. The SWAPPER process does not have quotas defined in
the same way other system and user processes do.
2The wait state specifies why a process cannot execute,
based on application-specific calculations.
|
