TOCPREVNEXTINDEX

Lund Performance Solutions


Performance Advice Message Catalog
Each of the advice messages follow a similar format. Advice messages are displayed on a terminal screen or in STDLIST in the case of a batch job. Messages can be sent to the console and to individual sessions via the TELLOP and TELL commands, respectively. All selected advice messages will be sent to a list of users and/or the system console. The TELL function is implemented as follows:
TELLOP
TELL MANAGER.SYS
TELL MGR.FINANC
<GI01> The CPU was...
Simply list potential target users by placing TELL or TELLOP in the first part of the file. To select specific messages to be sent, place an exclamation point (!) before the item. To select specific exceeded threshold messages, place an exclamation before the threshold.

Global Advice Messages <GXnn>

Each of the global messages relate to the "big picture" on the system. They will primarily refer to CPU states, but other events will be included as well. Global messages will include Informational (I) and Excessive (E) messages. "E" messages may require immediate action and should always be heeded.
<GI01> Global System CPU Usage
This informational message is a summary of the total amount of work actively performed by the CPU during the current interval. This number will be the same as the Total Busy: value in the Global CPU Statistics (tabular) section. Therefore, the CPU Statistics section does not have to be enabled in order to view this statistic.
<GI02> Process CPU Usage by Subqueue
This reveals the amount of CPU capacity being consumed within each subqueue. Noting the difference in utilization between the CS subqueue and the DS/ES subqueues, as this denotes how much CPU time is spent on interactive vs. batch processing. If the A and B subqueues are receiving an abundance of CPU time, this may indicate that system processes (or user processes that have "queue jumped") are creating problems. Unless special applications require the use of the B subqueue, most processing will occur in the CS subqueue during primary shift hours, and the DS subqueue during off-shift hours. This value is also found in the Global CPU Statistics section of the Global screen.
<GE01> Global System CPU Overhead Usage
This is an excessive condition indicator. It will only appear when the CPU Overhead percentage exceeds certain thresholds. This value is the same as the ICS/OH value in the CPU Global Statistics portion of the Global tabular screen. The default thresholds and their associated messages are listed below.
If CPU Overhead is greater than or equal to 10% and less than 12%:
CPU consumption due to system overhead during this interval was MODERATE
If CPU Overhead is greater than or equal to 12% and less than 14%:
CPU consumption due to system overhead during this interval was HEAVY
If CPU Overhead is greater than 14%:
CPU consumption due to system overhead during this interval was EXCESSIVE
As mentioned before, this statistic represents time spent by the CPU handling interrupt activity from DTCs and disc drives. Pressing RETURN (or ENTER) to get and MPE prompt is one such interrupt. Handling disc I/O completions are another.
The Advice section provides a message to help narrow down the cause(s) of high overhead utilization. Excessive terminal or disc I/O by a single (or multiple) process(es) can induce elevated overhead values and can affect response times significantly. This indicator is worth watching. If this number reaches as high as 30%, it is possible that a device (perhaps a modem or multiplexing device) is malfunctioning and is sending an inordinate number of interrupts to the system.
<GE02> Native Mode to Compatibility Mode Switch Rate
If this type of mode switching becomes excessive, the Advice section will display the appropriate message. Elevated mode switches can drain the CPU, forcing other processes to take longer to complete. This is especially true for NM to CM switches (see "CPU CM%"). These categories and the associated messages are listed below.
If the NM to CM switch rate is greater than or equal to 50 per second but less than 100:
Native Mode to Comp. Mode Switch rate during this interval was MODERATE
If the NM to CM switch rate is greater than or equal to 100 per second but less than 200:
Native Mode to Comp. Mode Switch rate during this interval was HEAVY
If the NM to CM switch rate is greater than 200 per second:
Native Mode to Comp. Mode Switch rate during this interval was EXCESSIVE
Refer to the Process Detail screen to find which processes are performing excessive switches.
<GE03> Compatibility Mode to Native Mode Switch Rate
If this type of mode switching becomes excessive, the Advice section will display the appropriate message. While CM to NM switches are less CPU-intensive than NM to CM switches, they still put an extra load on the CPU which can affect completion times for other processes. These categories and the associated messages are listed below.
If the CM to NM switch rate is greater than or equal to 100 per second but less than 200:
Comp. Mode to Native Mode Switch rate during this interval was MODERATE
If the CM to NM switch rate is greater than or equal to 200 per second but less than 300:
Comp. Mode to Native Mode Switch rate during this interval was HEAVY
If the CM to NM switch rate is greater than 300 per second:
Comp. Mode to Native Mode Switch rate during this interval was EXCESSIVE
<GE04> Global Average Response Times
This message appears when average prompt response times exceed various thresholds on the system. These categories and the associated messages are listed below.
If response time is greater than or equal to 10 seconds but less than 15:
Global average response time during this interval was MODERATE
If response time is greater than or equal to 15 seconds but less than 20:
Global average response time during this interval was HEAVY
If response time is greater than 20 seconds:
Global average response time during this interval was EXCESSIVE
<GE05> CPU Queue Length
This message appears when the CPU queue length exceed certain thresholds. Excessive CPU requests indicate that the CPU is not adequate for the amount of processing requested of it, or that too many jobs are being allowed to run concurrently. These categories and the associated messages are listed below.
If the CPU queue length is greater than or equal to 5 but less than 10:
CPU queue length indicates a MODERATE CPU bottleneck
If the CPU queue length is greater than or equal to 10 but less than 15:
CPU queue length indicates a HEAVY CPU bottleneck
If the CPU queue length is greater than 15:
CPU queue length indicates an EXCESSIVE CPU bottleneck
<GE06> Logon Failure Due to Shortage of Resources
This message appears when a logon fails because of a shortage of resources (CPU, disc, etc.). Excessive logon failures indicate that resources may be inadequate due to program hangs or database locks.
<GE07> Logon Failure Due to Shortage of Output Devices
This message appears when a logon fails because output devices are excessively involved. Excessive output requests can overload the CPU and contribute to poor performance.
<GE08> Logon Failure Due to Shortage of Disc Space
This message appears when a logon fails because disc space is lacking. When this occurs, programs usually abort or fail. Disc space must be freed up to improve performance.

Disc Advice Messages <DXnn>

<DE01> CPU Wait for Disc
This message appears when the CPU pause for disc I/O value exceeds certain thresholds. Excessive disc I/O indicates a possible shortage of memory or data locality issues. These categories and the associated messages are listed below.
If the CPU wait for disc value is greater than or equal to 5 but less than 10:
Disc I/O indicator #1 (CPU Pause Disc) reveals a MODERATE I/O bottleneck
If the CPU wait for disc value is greater than or equal to 10 but less than 15:
Disc I/O indicator #1 (CPU Pause Disc) reveals a HEAVY I/O bottleneck
If the CPU wait for disc value is greater than 15:
Disc I/O indicator #1 (CPU Pause Disc) reveals an EXCESSIVE I/O bottleneck

Memory Advice Messages <MXnn>

These messages are prefixed with an M, but follow the same format as Global messages. The majority of these messages are EXCESSIVE in nature are presented when memory resources are being heavily burdened. If HEAVY or EXCESSIVE messages are common on the system, it is very likely that the system does not have enough memory for the amount of processes required of it. CPU, disc, and response time indicators can all be adversely affected by a shortage of memory.
To alleviate the memory load, do one or more of the following:
  • Reschedule or redesign memory hog processes.
  • Add memory.
  • Restrict the number of jobs that may start up at any one time.
  • Reschedule batch processes to run after primary shift hours.
  • <ME02> CPU Busy on Memory Management
    The amount of time the CPU spends dealing with memory paging activity is usually proportional to the amount of memory necessary. When the CPU spends more than a few percentage points of its time managing main memory, there may be a memory shortage. These categories and the associated messages are listed below.
    If memory manager percentage is greater than or equal to 4% but less than or equal to 8%:
    Main Memory indicator #2 (CPU Memory) reveals a MODERATE memory load
    If memory manager percentage is greater than or equal to 9% but less than or equal to 12%:
    Main Memory indicator #2 (CPU Memory) reveals a HEAVY memory load
    If memory manager percentage is greater than 12%:
    Main Memory indicator #2 (CPU Memory) reveals an EXCESSIVE memory load
    <ME03> The Ratio of Memory Swaps to Process Launches
    Each time a process is granted the CPU’s attention, the dispatcher has to decide whether or not all the necessary data for that process are present in main memory. If all the data are not present, the memory manager has to perform a disc I/O. One swap per 10 launches is a ratio od 0.10 swaps per launch. As this ration escalates, the system works harder to satisfy memory requests, without actually performing more productive work. Consequently, response times can increase, especially if this ration reaches around 0.5 or higher. These categories and the associated messages are listed below.
    If swaps to launch ratio is greater than or equal to 0.4 but less than 0.6:
    Memory indicator #3 (Swap/Launch) reveals a MODERATE memory load
    If swaps to launch ratio is greater than or equal to 0.6 but less than 0.8:
    Memory indicator #3 (Swap/Launch) reveals a HEAVY memory load
    If swaps to launch ratio is greater than 0.8:
    Memory indicator #3 (Swap/Launch) reveals an EXCESSIVE memory load
    <ME04> The Page Fault Rate
    This value represents the number of times per second that memory page faulting occurred. A page fault is counted when a process needs a memory object (code or data) that is absent from main memory. Any consistent value of more than 25 page faults per second is indicative of a possible memory bottleneck.
    Note, however, that the page fault rate depends on the size of the system. See Table 9.6 for a breakdown of these rates. the value of 25 is used as a benchmark.
    The page fault categories and the associated messages are listed below.
    If the page faults per second rate is greater than or equal to 10 but less than 15:
    Memory indicator #4 (Page Fault Rate) reveals a MODERATE memory load
    If the page faults per second rate is greater than or equal to 15 but less than 20:
    Memory indicator #4 (Page Fault Rate) reveals a MODERATE memory load
    If the page faults per second rate is greater than 20:
    Memory indicator #4 (Page Fault Rate) reveals an EXCESSIVE memory load

    Process Advice Messages <PInn>

    Process Advice messages are prefixed with a P because they describe various conditions relating to individual process activity. Many times, a global problem is induced by one or two "problem" processes. It is especially critical to pay attention to the high resource usage process messages (identifying the "Hog" process).
    <GI02> Process CPU Usage by Subqueue
    See "<GI02> Process CPU Usage by Subqueue".
    <PI02> The Hog Process of the Current Interval
    The Hog process is simply the highest CPU-user for the current interval. There will always be an advice message indicating which process this is, even if the Hog is only using a tiny portion of the CPU. If, however, the Hog is consistently using vast amounts of CPU resource, you can drill down into this process by pressing the HOG PROC ZOOM function key. This one process may be affecting the performance of all other processes on the system.
    <PI03> The High Disc I/O Usage Process
    This message identifies the job or session that generated the highest number of combined reads and writes to disc during the current interval. If an extremely high number of disc I/Os are being performed by a particular process, it is also likely using a large portion of the CPU, which has to service the I/O. This message can indicate a number of issues at work. It may be that one or more TurboIMAGE datasets that were accessed by that process are experiencing data locality issues. This is an excellent method for finding application or data file inefficiencies.
    <PI04> The High Terminal I/O Process
    This message identifies the user session during the current interval that generates the highest number terminal reads. Refer to the "Global Misc Statistics (tabular format)" for more information on terminal I/O specifics. The high terminal I/O user will indicate one of three possibilities:
  • The user is holding the RETURN/ENTER key down.
  • The application is inefficient.
  • The application is efficient, but demanding. Some character mode applications generate terminal reads excessively.
  • If this process fall into one of the Application Workload groups that has been previously defined as having a certain amount of terminal reads per user transaction, then the number reported will reflect the approximate user transactions.

    Lund Performance Solutions
    www.lund.com
    Voice: (541) 812-7600
    Fax: (541) 812-7611
    info@lund.com
    TOCPREVNEXTINDEX