Aaron Grady [MSFT]
Guest
|
Posted:
Thu Jul 21, 2005 8:52 pm Post subject:
Re: Performance Counters |
|
|
Jairo-
Which part of the product are particularly interested in? The SNA gateway,
Transaction Integrator? I would agree there is unfortunately a lack of
information on this subject in the current documentation.
Here's some old information about some of the SNA Server counters. If you
can tell me which are you are interested I can see if I can dig anything
else up:
SNA Connections: Throughput Bytes/Sec
SNA Logical Unit Sessions: Throughput Bytes/Sec
SNA Adapter adaptername: Throughput Frames/Sec
These counters provide an indication of SNA activity. When observing these
counters, it may also be useful to start SNA Server Manager and double-click
on the server (the same one being observed in Performance Monitor). This
will allow you to see the number of users and sessions that correlate with a
particular level of SNA activity. This information, combined with data about
the processor and memory load, can help you understand the load and
performance on your servers. Low throughput does not necessarily mean poor
performance, but instead may simply indicate that current activity is low.
Measurement of frames/second may provide a better indicator of server load
than bytes per second, because server overhead (for interrupt handling and
message processing) increases per frame, not per byte. That is, a large
frame with many bytes requires about the same overhead as a small frame with
fewer bytes.
For suggested ways to tune SNA Server to maximize communications throughput,
see "Options for Maximizing Communications Throughput."
SNA Connections: Data Bytes Received/Sec
SNA Connections: Data Bytes Transmitted/Sec
SNA Logical Unit Sessions: Data Bytes Received/Sec
SNA Logical Unit Sessions: Data Bytes Transmitted/Sec
SNA Adapter adaptername: Data Bytes Received/Sec
SNA Adapter adaptername: Data Bytes Transmitted/Sec
SNA Adapter adaptername: Frames Received/Sec
SNA Adapter adaptername: Frames Transmitted/Sec
SNA Adapter adaptername: Throughput Bytes/Sec
These counters provide additional detail about SNA activity when used with
the previous three counters.
SNA Adapter adaptername: Adapter Failures
SNA Adapter adaptername: Connection Failures
SNA Adapter adaptername: Successful Connects
These counters may be useful for detecting patterns in which connections or
adapters fail for short periods and then return to normal. (Event logs can
provide more information about causes of failure.) You might also want to
set up Performance Monitor alerts with these counters, so that an alert is
triggered if too many failures are occurring.
HTH,
--
Aaron Grady
MCSE + Internet, MCDBA, MCSD, MCAD.NET
SNA Server/Host Integration Server
Microsoft
Blog: http://blogs.msdn.com/aagrady
This posting is provided "AS IS" with no warranties, and confers no rights.
Use of any included samples are subject to the terms specified at
http://www.microsoft.com/info/cpyright.htm
"Jairo Ammirati" <JairoAmmirati@discussions.microsoft.com> wrote in message
news:0A588A0D-5C31-4741-A45A-4CA8AD3361F4@microsoft.com...
| Quote: | Does anyone has a detail information about each counter that PerfMon
provide
for Host Integration?
I try to search but I only found incomplete information.
|
|
|
Jairo Ammirati
Guest
|
Posted:
Fri Jul 22, 2005 4:52 pm Post subject:
Re: Performance Counters |
|
|
Aaron,
thanks for your answer. But I have some counters here like this
% response < threshold 1
% response < threshold 2
% response < threshold 3
% response < threshold 4
% response > threshold 4
and I want to know what information and when I need to use this.
Thanks
"Aaron Grady [MSFT]" wrote:
| Quote: | Jairo-
Which part of the product are particularly interested in? The SNA gateway,
Transaction Integrator? I would agree there is unfortunately a lack of
information on this subject in the current documentation.
Here's some old information about some of the SNA Server counters. If you
can tell me which are you are interested I can see if I can dig anything
else up:
SNA Connections: Throughput Bytes/Sec
SNA Logical Unit Sessions: Throughput Bytes/Sec
SNA Adapter adaptername: Throughput Frames/Sec
These counters provide an indication of SNA activity. When observing these
counters, it may also be useful to start SNA Server Manager and double-click
on the server (the same one being observed in Performance Monitor). This
will allow you to see the number of users and sessions that correlate with a
particular level of SNA activity. This information, combined with data about
the processor and memory load, can help you understand the load and
performance on your servers. Low throughput does not necessarily mean poor
performance, but instead may simply indicate that current activity is low.
Measurement of frames/second may provide a better indicator of server load
than bytes per second, because server overhead (for interrupt handling and
message processing) increases per frame, not per byte. That is, a large
frame with many bytes requires about the same overhead as a small frame with
fewer bytes.
For suggested ways to tune SNA Server to maximize communications throughput,
see "Options for Maximizing Communications Throughput."
SNA Connections: Data Bytes Received/Sec
SNA Connections: Data Bytes Transmitted/Sec
SNA Logical Unit Sessions: Data Bytes Received/Sec
SNA Logical Unit Sessions: Data Bytes Transmitted/Sec
SNA Adapter adaptername: Data Bytes Received/Sec
SNA Adapter adaptername: Data Bytes Transmitted/Sec
SNA Adapter adaptername: Frames Received/Sec
SNA Adapter adaptername: Frames Transmitted/Sec
SNA Adapter adaptername: Throughput Bytes/Sec
These counters provide additional detail about SNA activity when used with
the previous three counters.
SNA Adapter adaptername: Adapter Failures
SNA Adapter adaptername: Connection Failures
SNA Adapter adaptername: Successful Connects
These counters may be useful for detecting patterns in which connections or
adapters fail for short periods and then return to normal. (Event logs can
provide more information about causes of failure.) You might also want to
set up Performance Monitor alerts with these counters, so that an alert is
triggered if too many failures are occurring.
HTH,
--
Aaron Grady
MCSE + Internet, MCDBA, MCSD, MCAD.NET
SNA Server/Host Integration Server
Microsoft
Blog: http://blogs.msdn.com/aagrady
This posting is provided "AS IS" with no warranties, and confers no rights.
Use of any included samples are subject to the terms specified at
http://www.microsoft.com/info/cpyright.htm
"Jairo Ammirati" <JairoAmmirati@discussions.microsoft.com> wrote in message
news:0A588A0D-5C31-4741-A45A-4CA8AD3361F4@microsoft.com...
Does anyone has a detail information about each counter that PerfMon
provide
for Host Integration?
I try to search but I only found incomplete information.
|
|
|
Aaron Grady [MSFT]
Guest
|
Posted:
Wed Jul 27, 2005 4:54 pm Post subject:
Re: Performance Counters |
|
|
Ah.... this is Response Time Monitoring stuff. Here's a big excerpt from one
of the resource kits on RTM:
Response Time Monitor
During 3270 emulation, SNA Server maintains statistics on the time it takes
for the host to respond to a 3270 client request. Response Time Monitor
(RTM) data can be sent to the host for collection and analysis; the 3270
user can also view a graphical display of the spread of response times on a
particular 3270 emulation session. The status line at the bottom of the
screen can also display the time for the last transaction executed.
Use the SNA Server management console to choose the way in which response
times are measured and classified. RTM statistics for a particular LU
(logical unit) are sent to the host that owns that LU rather than sent on
the connection designated for NetView data.
You can specify the times at which RTM should send data and also the trigger
that will cause RTM to register that the host has responded. Configuration
settings are on the Response Time Monitor (RTM) tab of the Properties page
for the selected SNA Server subdomain. You can configure the settings shown
in the following table:
--------------------------------------------------------------------------------
Note For these settings to be meaningful, each emulator for your 3270 users
must support RTM.
--------------------------------------------------------------------------------
Table 7.1 RTM Settings
This setting Can be configured to
RTM Data Sent At Specify when RTM data is sent to the host. You can
send the data during one or both of the following situations:
Counter Overflow Send the RTM data to the host when the number of
host responses in a given time period overflows the size of the available
counter.
End of Session Send the RTM data to the host at the end of each
LU-to-LU session.
RTM Timers Run Until Set the point at which RTM will register that a
host has responded; this is when RTM stops the timers. (The timers are
started when the local system sends data.) The possible stopping points are
as follows:
First Data Reaches Screen Stop timing when data reaches the local
screen.
Host Unlocks Keyboard Stop timing when the host unlocks the local
keyboard.
Host Lets User Send Stop timing when the host lets the local computer
send more data.
RTM Thresholds Set the cutoff times, in tenths of a second, at which
RTM saves its count of host responses and then restarts the count. For
example, you could specify 5, 10, 20, and 50 to save the count of host
responses during the intervals from 0.0 to 0.5 seconds, from 0.5 to 1.0
seconds, from 1.0 to 2.0 seconds, and from 2.0 to 5.0 seconds.
The range is from 1 through 1000 in tenths of a second (from 0.1
seconds to 100.0 seconds). The defaults are 5, 10, 20, and 50 (0.5 seconds,
1.0 seconds, 2.0 seconds, and 5.0 seconds).
--------------------------------------------------------------------------------
Note If the host RTM settings are different, they override the SNA Server
settings.
--------------------------------------------------------------------------------
Collecting RTM Data and Sending It to the Host
SNA Server always collects RTM data while 3270 emulation is active. The data
for each 3270 LU is then sent to the host to which the LU is connected under
various criteria.
You configure the initial criteria for SNA Server to send RTM data to the
host when any of the RTM counters reaches its maximum value and/or each time
a 3270 session ends. (For instructions on configuring the RTM function, see
the Microsoft SNA Server Administration Guide.) The RTM counters are reset
each time the RTM data is sent in this way.
Although initial criteria apply to all SNA Server LUs, any host can change
the criteria for any or all LUs it owns, and solicit or reset any LU
counters.
Boundaries and Measurement Definition
RTM data is collected by comparing host response times against a series of
four boundary values. Each time a host transaction occurs, the response time
is compared with the boundary values, and the appropriate counter is
incremented. There is a counter for each of the four intervals defined by
the boundary values, and an overflow counter for response times above the
largest boundary value.
You can change the default boundaries - 0.5, 1, 2, and 5 seconds - using the
SNA Server management console. You also can select which of three response
time definitions to set as default.
The response time is measured from the time the user presses ENTER until one
of the following events occurs:
a.. The first character of host data reaches the 3270 display.
b.. The keyboard is unlocked.
c.. The user is allowed to send data.
Again, a host can override the default boundaries and other host response
time settings for any or all LUs it controls. (For more information, see
"Host Modification of RTM Parameters" later in this chapter.)
Local Display
You can use the SNA Server management console to allow a 3270 user to view
the last transaction time on the status line and a graphical display of the
response times for each 3270 display session. For instructions, see the
Microsoft SNA Server Administration Guide. Again, the host can override this
ability to view the RTM data locally.
Reuse of LUs by Different Users
Response times are recorded for a specific LU rather than for a specific
3270 session. This means that if a 3270 session ends and another session (of
the same or a different 3270 user) then uses the same LU, the response time
counters include responses from the previous session as well as those from
the new session. This does not happen, however, if the RTM counters were
reset before the new session began to use the LU. (The counters are reset
when the host requests that they be reset or when RTM data is sent
unsolicited by SNA Server). When a host requests RTM data, the host also
typically requests that the counters be reset.
If you want to ensure that the RTM data displayed relates only to the
current 3270 session, you should select the RTM Data Sent at End-of-Session
checkbox when configuring RTM parameters for SNA Server. This will ensure
that the RTM counters are reset at the end of each 3270 session and will
therefore be 0 at the start of each 3270 session. The host can, of course,
override this setting.
Host Modification of RTM Parameters
Although you can configure the RTM boundaries and definition that you want
to use on your server, the host can override these configured values, either
for an individual 3270 LU or for all LUs it controls. The host can also
specify whether to permit local display of RTM data, and, when SNA Server
sends RTM data, it can disable collection of RTM statistics completely.
Lost Data Indicator
When SNA Server sends RTM data to the host, it can indicate to the host that
RTM data may have been lost (that is, some host response times were not
included in the RTM data sent to the host). This indication is sent in the
following cases:
a.. One of the RTM counters reaches its maximum value, and SNA Server is
not configured to send RTM data unsolicited at counter overflow. When this
occurs, SNA Server stops recording response times until the host requests
the data (or until end-of-session, if unsolicited sending at end-of-session
is configured). The data is sent with an indication that data may have been
lost due to counter overflow.
b.. When the local server terminates abnormally, RTM data obviously is
lost and any connection to the host is terminated. When the connection to
the host is started, there is no way of knowing whether the connection was
previously terminated by a server failure. Therefore, the first RTM data
sent to the host after the connection is established includes the lost-data
indicator to indicate that data may have been lost due to failure of the
local server.
c.. When SNA Server is configured to send RTM data at end-of-session, and
a new 3270 session is started before the RTM data for the previous 3270
session on the same LU can be sent, response times for the new session are
discarded until the RTM data from the previous session is sent and the
counters are reset. When RTM data is next sent on the new session, it
includes the lost data indicator to indicate this condition.
--
Aaron Grady
MCSE + Internet, MCDBA, MCSD, MCAD.NET
SNA Server/Host Integration Server
Microsoft
This posting is provided "AS IS" with no warranties, and confers no rights.
Use of any included samples are subject to the terms specified at
http://www.microsoft.com/info/cpyright.htm
"Jairo Ammirati" <JairoAmmirati@discussions.microsoft.com> wrote in message
news:DA14705F-0A02-4D9A-AA92-596AB5C14D0F@microsoft.com...
| Quote: | Aaron,
thanks for your answer. But I have some counters here like this
% response < threshold 1
% response < threshold 2
% response < threshold 3
% response < threshold 4
% response > threshold 4
and I want to know what information and when I need to use this.
Thanks
"Aaron Grady [MSFT]" wrote:
Jairo-
Which part of the product are particularly interested in? The SNA
gateway,
Transaction Integrator? I would agree there is unfortunately a lack of
information on this subject in the current documentation.
Here's some old information about some of the SNA Server counters. If you
can tell me which are you are interested I can see if I can dig anything
else up:
SNA Connections: Throughput Bytes/Sec
SNA Logical Unit Sessions: Throughput Bytes/Sec
SNA Adapter adaptername: Throughput Frames/Sec
These counters provide an indication of SNA activity. When observing
these
counters, it may also be useful to start SNA Server Manager and
double-click
on the server (the same one being observed in Performance Monitor). This
will allow you to see the number of users and sessions that correlate
with a
particular level of SNA activity. This information, combined with data
about
the processor and memory load, can help you understand the load and
performance on your servers. Low throughput does not necessarily mean
poor
performance, but instead may simply indicate that current activity is
low.
Measurement of frames/second may provide a better indicator of server
load
than bytes per second, because server overhead (for interrupt handling
and
message processing) increases per frame, not per byte. That is, a large
frame with many bytes requires about the same overhead as a small frame
with
fewer bytes.
For suggested ways to tune SNA Server to maximize communications
throughput,
see "Options for Maximizing Communications Throughput."
SNA Connections: Data Bytes Received/Sec
SNA Connections: Data Bytes Transmitted/Sec
SNA Logical Unit Sessions: Data Bytes Received/Sec
SNA Logical Unit Sessions: Data Bytes Transmitted/Sec
SNA Adapter adaptername: Data Bytes Received/Sec
SNA Adapter adaptername: Data Bytes Transmitted/Sec
SNA Adapter adaptername: Frames Received/Sec
SNA Adapter adaptername: Frames Transmitted/Sec
SNA Adapter adaptername: Throughput Bytes/Sec
These counters provide additional detail about SNA activity when used
with
the previous three counters.
SNA Adapter adaptername: Adapter Failures
SNA Adapter adaptername: Connection Failures
SNA Adapter adaptername: Successful Connects
These counters may be useful for detecting patterns in which connections
or
adapters fail for short periods and then return to normal. (Event logs
can
provide more information about causes of failure.) You might also want to
set up Performance Monitor alerts with these counters, so that an alert
is
triggered if too many failures are occurring.
HTH,
--
Aaron Grady
MCSE + Internet, MCDBA, MCSD, MCAD.NET
SNA Server/Host Integration Server
Microsoft
Blog: http://blogs.msdn.com/aagrady
This posting is provided "AS IS" with no warranties, and confers no
rights.
Use of any included samples are subject to the terms specified at
http://www.microsoft.com/info/cpyright.htm
"Jairo Ammirati" <JairoAmmirati@discussions.microsoft.com> wrote in
message
news:0A588A0D-5C31-4741-A45A-4CA8AD3361F4@microsoft.com...
Does anyone has a detail information about each counter that PerfMon
provide
for Host Integration?
I try to search but I only found incomplete information.
|
|
|