Windows Performance Monitor
To get a good overview over the performance a given system can deliver, the ‘Windows Performance Monitor’ is a pretty handy tool. It can give a real time graph of specific performance specifications or throughput, so called performance counters. These counters are tracked in specified intervals and may be logged in a file for later review.
In general, there are 4 “pillars” of performance in a system:
- Disk access/throughput
For each of these topics, there exist predefined counters inside the performance monitor.
The ‘Windows Performance Monitor’ is located inside the ‘Control Panel’, under ‘Administrative Tools’:
When you open it, you see a real time graph with some counters which are set per default:
You can add counters to this real time view, by clicking on the “PLUS” button above the graph:
In the upcoming dialogue, you can select the machine, for which you would like to add the counters, as well as the counters itself:
At the top, you select the machine, either the local one or a remote machine.
Below, you have to select a specific performance object, which is more or less a category for the counters. In the screenshot above, the Processor object is selected, so in the list below you can select performance counters, which fall under the category processor.
You can either select specific counters or all counters from the selected object. In addition, you may specify the instance of the object which should be logged. In the case of Processor object, these are the different CPU (Cores) of a machine. You can specifically select an instance, or simply select all. Click “Add” to add the counter to the current log.
As described above, general performance can be measured in 4 categories:
- Disk access/throughput
For each of these categories, objects and counters exist in the system.
Format is <Object>/<Counter>, normally valid for “All Instances”
Processor / % Processor Time
Process / % Processor Time (to view specific processes)
Memory / Available Kbytes
PhysicalDisk / % Idle Time (to get busy time of the disk, simply calculate 100 - % Idle Time) PhysicalDisk / Disk writes/sec
PhysicalDisk / Disk reads/sec PhysicalDisk / Avg. Disk Queue Length
Network Interface / Bytes sent/sec (for the specific NIC)
Network Interface / Bytes received/sec (for the specific NIC)
For TCOSS, some additional objects are installed.
The most important ones are:
- TCOSS Disk
- TCOSS Links
For these, you should select all counters.
Creating a log for further analysing
Normally, the performance monitor gives you a real-time view of the currently defined counters. This, however, gives only an overview over a recent short timeframe, not over a longer period of time.
Therefore it is possible, to generate a counter log, which creates a log file, in which the counters are stored.
To start such a log file click on the “Counter Logs” section in “Performance Logs and Alerts”.
On the right hand side, you can see your currently defined logs, and if they are active or not. Normally you should only see the “System Overview” here. In the screenshot below there is also a log called “TCPERFLOG”. Please see details on this a little further below, in the “Other Tools” section of this article.
To create a new log, click on the empty page button.
Here you have to give it a Name, for example “Performance Log”
In the following window, you can define the properties of the log, such as the defined counters, or location and behaviour of the log files itself.
You can either add a specific counter, like in the ‘Live Monitor’ above, or add complete Objects. Click on ‘Add Objects’ and add the “Processor” Performance Object for the local machine to the log.
Here you also have the possibility to click the “Explain” button, which gives a small description of the selected ‘Performance Object’.
Unfortunately, this does not work with Counters created by KCS products. For descriptions on these, please see the TCOSS System Manual, chapter “1.13 - Performance Counters”.
After clicking “Add”, click “Close” to go back to the properties page of the log.
Below, you can set the interval, in which the counters should be read out and posted to the log, ranging from once per second to once every couple of days.
On the tab “Log File” you can set the location and name of the log, as well as the type of logging, max size and log file numbering.
Normally, you will either use a binary or a binary circular log. The first one starts at a specified time and logs until it is stopped, or it has reached its maximum size.
The circular log behaves basically the same, but it will not stop when it reaches max size. Instead, the oldest entries are deleted, each time a new entry is made.
If you want to log during a specific timeframe, a binary log with specified start and end times may be the best solution, as you do not have the risk of overwriting data.
However, if you want an overview over the recent performance of a running system and intend to log constantly, then a circular log is the better idea.
You can now send this newly generated log file to KOFAX Vienna BC Technical Support for further analyzing. Always in combination with a problem ticket of course.
To not always have to restart the logging in the performance monitor after a reboot of the machine, there is the TC/PerfLog tool. This tool runs under the TCSRV service and is controllable inside TC Monitor.
This tool is part of the TC/SP package.
TC/PerfLog Manual excerpt:
“TC/PerfLog is a component of KCS Server Package that controls the process of collecting + logging of performance counters: once the parameters for performance counter logging defined, TC/PerfLog provides permanent logging (e.g. automatic log-restart after machine reboot) and log file size control (=keeping only the most recent log files according to the LogSizeTotalMax setting) .“
For installation and configuration details, please see the TC/PerfLog Manual.
TC/PerfLog also comes with a predefined counter set, which has the most important counters already defined.
After installing TC/PerfLog, in the “Counter Log” list in the Performance Monitor, the TCPERFLOG counter log has been added to the list.
It is a circular log file, stored under C:\TCOSS\PERFLOG\ per default. All settings for this log are either to be configured in the registry, or via one of 2 text files, also stored under C:\TCOSS\PERFLOG.
For configuration details, please also have a look in the TC/PerfLog Manual.
TCDiskTest is a tool, which simulates TCOSS read and write operations and provides performance counters to measure the access time and throughput. The intention is to provide this tool to a new customer to evaluate the HDD performance of the servers planned for TCOSS installation, before really doing the installation.
This tool can also be found in the TC/SP package, but is not part of the setup routine. It can be found in the win32 subdirectory and only has a command line interface.
For information more on using this tool, please have a look in the TCOSS Platform Manual, Chapter 5.4.2 Checking Disk IO Latency.
In the TC/SP documentation, there exists an Excel spreadsheet, called “HW- RequirementCalculator_XXXXX.xlsx” where “XXXXX” is the version number of the package the sheet is taken from.
In the sheet, “inbound” means FAX -> TCOSS -> Link -> Mail system, “outbound” means Mail system -> Link -> TCOSS -> FAX.
Here you can enter the desired specs of your TCOSS system. These have to be entered in the yellow fields.
Number of fax lines
How many lines are dedicated to in or outbound traffic on the system, at the same time
Keep log file
Generate journal files or not
Route via in-action
If and how many in-events are triggered for an arriving fax
Sending with back reception
If back reception is generated when sending faxes
Sending with delivery notification
If a delivery notif is requested for sending
Average message size in bytes
The average message size of the fax in TCOSS, (mostly test, or depends on the type of faxes sent with pictures etc.)
Fax line usage per message in seconds
number of seconds the fax line is occupied receiving the average inbound/outbound fax, including idle time before the next message is received/sent
Average number of pages
The average number of pages for faxes sent/received
Peak message throughput per hour
This is calculated autmatically for fax lines; however, it has to be given for a connected link. Inbound it is roughly the same as fax throughput (normally all received faxes are passed on to a lin). Outbound, the link puts the messages faster inside the fax queue (practically about a factor of 2)
The blue fields at the bottom tell you what specifications your Hard Drive has to provide to support these specifications. To be on the safe side, you should add a buffer of 50-100% on both of these values, and be sure the system is not always on full load.
The “disk writes per second” can be matched with the performance counter “PhysicalDisk / Disk writes/sec”, but the “average disk access time” has to be taken from the manufacturer specs of your HDD:
For more information on this sheet, please have a look inside the TCOSS Platform Manual, chapter 6.4.2 KCS HW-Requirement Calculator.
Problems with virtual Memory and TCOSS
For TCOSS, in the registry you can set the size of caches used by the server process
|DirCacheSize||Integer||2048||Directory cache size in kB|
|DocCacheSize||Integer||40960||Document cache size in kB|
|DataBaseCacheSize||Integer||5120||Database cache size in kB|
|CacheAweEnable||Integer||0||Cache AWE (Address Windowing Extensions) enable (1 = enable)|
These caches will be reserved by the TCOSS process in the “virtual bytes”. This resource can be viewed via the performance counter ‘Process\Virtual Bytes’ – Instance.
These virtual bytes can have, depending on the system size, after a TCOSS start from 100 to 500 MByte in size. In addition to this start value, the cache size is also added there directly.
The problem here is, this value is not allowed to exceed 2GByte on 32bit systems and 4GByte on 64bit systems, otherwise TCOSS will not start anymore.
- Microsoft Windows Server 2003 32-bit x86
- Microsoft Windows Server 2008, 32-bit or 64-bit
Keywords: Perfmon, counters,TCOSS, percentage,