h

IO

mon

™

  The File I/O Performance Monitor

Home  |  Products & Services  |  Contact  |  About hyperI/O^sm

Big Picture

hIOmon SSD I/O Performance Metrics Analysis
 Add-On Support

Product Info

The hIOmon "Solid State Disk (SSD) I/O Performance Analysis Add-On" enables you to quickly and easily configure the hIOmon software so as to collect a comprehensive set of "Device Summary" I/O operation metrics that reflect not only the actual usage of SSD-related TRIM commands but also actual read and write I/O operation activity for those specific physical devices and Logical Disks of particular interest to you.

Features

Architecture

Screen Shots

Documentation

Download

NOTE:

The I/O operation performance metrics collected through the use of the hIOmon "SSD I/O Performance Analysis" Add-On are empirical metrics that can reflect the actual I/O operations performed through the normal, everyday use of your very own applications and associated files!  And no file, application, or operating system changes are required!  

The configuration script along with the hIOmon Excel template/macros provide a quick and easy way to help you evaluate and understand the performance impact of SSDs, once again based upon empirical metrics directly related to the individual files and devices that you are actually using.  Additional information is provided below (and also within the "hIOmon Add-On User Guide" document).

An online tutorial that illustrates the "step-by-step" instructions described below is also available.

Also see the "hIOmon SSD TRIM Metrics Display Gadget" for a a simple and quick way to display a selected subset of the SSD TRIM-related metrics that can be captured by the hIOmon File I/O Performance Monitor.

hIOmon Configuration Setup for SSD I/O Operation Metrics

The hIOmon “Automated Monitoring Configuration Setup” VBScript provides a specific “SSD I/O Performance Analysis” option that uses the hIOmon WMI Support to help automatically define and activate a Filter Selection (that identifies which particular physical devices and optionally Logical Disks are to be monitored by hIOmon) and to automatically create and activate a "hIOmon Manager Export File" to which a specific set of "summary I/O operation" performance metrics are exported (saved).

1. That the hIOmon File I/O Performance Monitor software be properly installed and active (i.e., that the hIOmon Manager and hIOmon I/O Monitor components are currently running; note that these components are normally configured upon the installation of the hIOmon software to automatically be started when the computer system is started).

2. That both the "Summarized Data" and "Raw Device Extended Feature" options are licensed.  Note that the trial version of the hIOmon File I/O Performance Monitor software includes full-support for both of these licensing options.

3. This script is interactive; as such it requires limited user input (e.g., the specific Logical Disks for those particular physical devices that are to be monitored by hIOmon) as shown below

The "hIOmonAMCSS.vbs" VBScript file can be found in the "VBScript" folder within the "..\hIOmon\Support\hIOmonWMI\Scripting\Wsh" directory where the hIOmon software was installed.

The script can also be run by means of the "Start -> Programs -> hIOmon -> Getting Started Help -> hIOmon Automated Monitoring Configuration Setup" shortcut (that was added to the Start Menu by the Windows Installer as part of the hIOmon software installation process) or optionally run as part of the Windows Installer installation process for the hIOmon software.

To run the script as part of the Windows Installer installation process for the hIOmon software, simply check the "Invoke the hIOmon Automated Monitoring Configuration Setup Script" checkbox on the "Custom Installation Options" display that is shown as part of the installation process.

To run the script anytime after the hIOmon software has been installed, simply double-click upon the script file (or the shortcut above), which will then display a prompt requesting the particular configuration option to be performed. Select the "SSD I/O Performance Analysis" option as shown below, then click on the "OK" button:

The configuration script will subsequently display the following prompt asking whether you want to also collect both read and write I/O operation metrics in addition to the MDSA-related metrics for all of the physical devices to be subsequently specified.

To also collect read and write I/O operation metrics for all physical devices to be monitored, enter "1"; otherwise to collect MDSA-related metrics only, enter "0" (or simply click on the "Cancel" button: 

Automated Filter Selection creation

Once you have specified all of the Physical Devices (and optionally Logical Disks) that you want monitored by hIOmon, the script will automatically create a Filter Selection (named "SSDPerfAnalysisFS" by default), which identifies the specific devices for which I/O operation performance metrics are to be collected by hIOmon.  The script will also make this Filter Selection the “default” Filter Selection (so that it will automatically be used each time the hIOmon Manager service is started as part of starting your computer).

Automated hIOmon Manager Export File creation

After creating the Filter Selection, the configuration script will then display a prompt (shown below) requesting the name of the "hIOmon Manager Export File" to be created.  As it retrieves the summary device I/O operation performance metrics collected by the hIOmon I/O Monitor component, the hIOmon Manager will directly save/export a particular limited set of these metrics to the hIOmon Manager Export File.  And as described below, the specific set of metric types exported is dependent upon whether the hIOmon software has been configured to additionally collect read and write I/O operation metrics and to collect I/O operation performance metrics for Logical Disks.

In the event that a system reboot is not required, the script will automatically activate the Filter Selection (so that the hIOmon I/O Monitor will immediately begin collecting metrics for those devices specified within the Filter Selection) and then display the following message: 

The script will also automatically create a "hIOmon Manager Export File" (named "SSDPerfAnalysisExportFile" by default) that is to include the particular I/O operation performance metric types for which collected metrics are to be exported/saved.  The particular metric types to be exported are noted below.

The script will also automatically activate/start this hIOmon Manager Export File so that the hIOmon Manager can immediately begin exporting the collected I/O operation performance metrics to this export file.  The following message will be displayed by the script to confirm that exporting has been activated.  In addition, the hIOmon Manager Export File will be made a "default" export file (so that it will automatically be used each time the hIOmon Manager service is started as part of starting the computer system).  

That's it!  Simply enter the letters of the Logical Disks of the physical devices (and optionally of the Logical Disks themselves) for which you want the hIOmon software to collect I/O operation performance metrics (and reboot the system if necessary).  The configuration script automatically configures the hIOmon software as required based upon the Logical Disk letters that you specify.

As described below, this script can also be subsequently used to change the Filter Selection to include additional Physical Devices (and Logical Disks) to be monitored by hIOmon.  The hIOmon "Excel Template/Macros" description provides information on how you can import the I/O operation performance metrics contained with the hIOmon Manager Export File into an Excel worksheet and generate graphs/charts based upon these metrics.

Also please note that:

• This script and configuration option can be used during the installation of the hIOmon software (or anytime thereafter).

• If you select to monitor a Logical Disk, then the hIOmon software will monitor all files upon the specified Logical Disk.

• The I/O operation performance metrics will be collected for all of the monitored devices (both physical devices and any Logical Disks) upon an I/O "summary" basis for each individual monitored device.  These metrics will be collected for the associated device in aggregate (i.e., upon a "Device Summary" basis, which in the case of Logical Disks reflects the accumulated I/O operation performance metrics for all monitored files resident upon the same Logical Disk).

• "Physical Device Extended Metrics" will also be collected for all Logical Disks specified to be monitored by hIOmon.  These metrics basically reflect "physical device" I/O operations that were required to complete file I/O operations for those files that were also being monitored by the hIOmon I/O Monitor.

• The "summarized" metrics will be collected by the hIOmon I/O Monitor for offload/export upon a periodic basis (at 10 minute intervals by default unless a different interval period was explicitly specified).  Note also that by default only “Device Summary” metrics will actually be offloaded to the hIOmon Manager Export File (i.e., summary metrics for the individual files monitored are not offloaded/exported).

• The configuration script will automatically exclude the "..\hIOmon\Support\hIOmonGr" directory (where the hIOmon software was installed) from being monitored by hIOmon if this directory resides upon a Logical Disk that is being monitored by hIOmon.

Back to Top     

Simply double-click upon the script file (or on the shortcut above) to run the script again and select the "SSD I/O Performance Analysis" option.  Since the Filter Selection will have been created when the script was previously executed, the following message will be displayed: 

To add another Physical Disk (and/or Logical Disk) to the existing Filter Selection, simply click on the "Yes" button, which will cause the script to display the "Logical Disk letter" prompt for input (as previously shown above) along with the other input prompts as shown above.

To start with a new "SSDPerfAnalysisFS" Filter Selection, click on the "No" button, which will cause the script to delete the existing Filter Selection and then display the same input prompts as mentioned above so as to create a new Filter Selection.

Once you have completed modifying the Filter Selection, the script will then again determine whether the "physical disks" to be monitored by hIOmon are currently attached to the hIOmon I/O Monitor component.  If not, the script will display the message saying that a system reboot is required.

Click on the "Cancel" button to immediately end the script, in which case no hIOmon software configuration changes will be made.  

After you are done modifying the Filter Selection, the script will then try to create the hIOmon Manager Export File. The script will display the following message if it detects an existing hIOmon Manager Export File with the same name:

Click on the "OK" button to delete the existing hIOmon Manager Export File.  After this hIOmon Manager Export File has been deleted, the script will then create a new hIOmon Manager Export File (with the same name), which will then be activated and made a default by the script.

The script will subsequently display messages to indicate that the modified/new Filter Selection has been activated/loaded and that the new hIOmon Manager Export File has also been activated unless a system reboot is required.

The script automatically configures the hIOmon Manager Export File to include the following MDSA-related metric types:

• Name of the monitored device

• Timestamp (month, date, hour, minute) when the metrics were prepared for offload to the hIOmon Manager service (so that they could then be exported/saved to the export file)

• TRIM IOP Count (i.e., the number of successful Control I/O operations observed by the hIOmon I/O Monitor that specified a "Device Control" operation with a MDSA request for a "TRIM" action)

• TRIM DSR Total Count (i.e., the total combined number of "Data Set Ranges (DSR)" specified by the TRIM commands.  This is the cumulative total for all of the TRIM requests as indicated by the "TRIM IOP Count" value above.)

• TRIM DSR Total Length (i.e., the total combined length (in bytes) of the DSRs specified by the TRIM commands.  This is the cumulative total for all of the TRIM requests as indicated by the "TRIM IOP Count" value.)

• TRIM DSR Length Max (i.e., the overall maximum length (in bytes) of a DSR specified by the TRIM commands.  This is the largest length observed for all of the TRIM requests as indicated by the "TRIM IOP Count" value).

• TRIM DSR Length Min (i.e., the overall minimum length (in bytes) of a DSR specified by the TRIM commands.  This is the smallest length observed for all of the TRIM requests as indicated by the "TRIM IOP Count" value.)

• TRIM DSR Offset Max (i.e., the maximum (i.e., highest) starting address (in bytes) of a DSR specified by the TRIM request as observed by the hIOmon I/O Monitor)

• TRIM DSR Offset Min (i.e., the minimum (i.e., lowest) starting address (in bytes) of a DSR specified by the TRIM request as observed by the hIOmon I/O Monitor)

• TRIM IOP DSR Max Count (i.e., the maximum total number of DSRs specified by a single TRIM command as observed by the hIOmon I/O Monitor.  That is, the combined total number of Data Set Ranges specified by a single TRIM request)

• TRIM IOP DSR Min Count (i.e., the minimum total number of DSRs specified by a single TRIM command as observed by the hIOmon I/O Monitor.  That is, the combined total number of Data Set Ranges specified by a single TRIM request)

• TRIM IOP DSR Max Length (i.e., the maximum total combined lengths (in bytes) of the DSRs specified by a single TRIM command as observed by the hIOmon I/O Monitor.  That is, the maximum total number of bytes specified by a single TRIM request.)

• TRIM IOP DSR Min Length (i.e., the minimum total combined lengths (in bytes) of the DSRs specified by a single TRIM command as observed by the hIOmon I/O Monitor.  That is, the minimum total number of bytes specified by a single TRIM request.)

• Duration (i.e., the length of time in seconds during which the associated summary I/O operation performance metrics were collected by the hIOmon I/O Monitor)

• Statistics Time Duration Base (i.e., time duration in seconds since the hIOmon I/O Monitor first started accumulating the associated summary I/O operation performance metrics)

If read and write I/O operation metrics were specified to also be collected for the monitored devices, then the script automatically configures the hIOmon Manager Export File to additionally include the following I/O operation performance metric types for the monitored physical devices (and any monitored Logical Disks):

• Separately for both read and write I/O operations and as observed by the hIOmon I/O Monitor:
  -  Accumulated number of I/O operations
  -  Accumulated number of "Fast" I/O operations; that is, I/O operations that were successfully completed in less than one millisecond as observed by the hIOmon I/O Monitor either by explicit request for use of the "system file cache" (SystemCache Hits) or not (i.e., successfully completed in less than one millisecond but did not explicitly request the use of the "system file cache")
  -  Accumulated number of "System File Cache" I/O operations; that is, I/O operations that were successfully performed using the system file cache by explicit request as observed by the hIOmon I/O Monitor
  -  Accumulated number of System File Cache "miss" I/O operations; that is, I/O operations that were successfully performed using the system file cache by explicit request, but took one or more milliseconds to complete as observed by the hIOmon I/O Monitor.  Note that the "System Cache" I/O operation "hit" count can be derived by subtracting the "miss" count from the total System Cache I/O operation count (i.e., the "System File Cache" I/O operation count defined above); this derived "hit" count represents the number of I/O operations that were successfully performed using the system file cache by explicit request and which completed within less than one millisecond as observed by the hIOmon I/O Monitor.
  -  Average queue length (i.e., "outstanding I/O operations")
  -  Maximum queue length
  -  Accumulated total (in seconds) of the time durations associated with the I/O operations
  -  Maximum time duration (in seconds) for an observed I/O operation (i.e., maximum response time)
  -  Accumulated total (in bytes) of data transferred by the I/O operations
  - Largest data transfer size (in bytes) associated with an observed I/O operation (i.e., maximum data transfer size)
  -  Accumulated number of I/O operations that represent random access (i.e., the random access I/O count).  An I/O operation is considered to be a random access when the data transferred at its starting address is not contiguous to the address of the last data transferred by the preceding I/O operation.
  -  Accumulated total (in bytes) of data transferred by I/O operations that performed a random access (i.e., the random access data transferred amount)
  -  Accumulated number of I/O operations that represent sequential access (i.e., the sequential access I/O count).  An I/O operation is considered to be a sequential access when the data transferred at its starting address is contiguous to the address of the last data transferred by the preceding I/O operation.
  -  Accumulated total (in bytes) of data transferred by I/O operations that performed a random access (i.e., the sequential access data transferred amount)
  -  Accumulated total (in bytes) of data transferred by I/O operations that were successfully performed using the system file cache by explicit request and which completed in less than one millisecond as observed by the hIOmon I/O Monitor (i.e., the system file cache hit data transferred amount)
  -  Accumulated total (in bytes) ) of data transferred by I/O operations that were successfully performed using the system file cache by explicit request, but which took one or more milliseconds to complete as observed by the hIOmon I/O Monitor (i.e., the system file cache miss data transferred amount)

• Average and maximum queue length for read and write I/O operations combined

• Accumulated number of I/O operations (whether Read or Write) that represented random accesses to the file regardless of the particular open file instance. In this case an I/O operation is considered to be a random access when the data transferred at its starting file address is not contiguous to the file address of the last data transferred by the preceding I/O operation and regardless of whether the preceding I/O operation was directed to another open instance of the same file.

• Accumulated number of I/O operations (whether Read or Write) that represented sequential accesses to the file regardless of the particular open file instance. In this case an I/O operation is considered to be a sequential access when the data transferred at its starting file address is contiguous to the file address of the last data transferred by the preceding I/O operation regardless of whether the preceding I/O operation was directed to another open instance of the same file

• Idle time total (i.e., accumulated total duration in seconds as observed by the hIOmon I/O Monitor in which there were neither I/O operations in process nor any I/O operations queued)

If the Filter Selection includes Logical Disks to be monitored, then the script automatically configures the hIOmon Manager Export File to additionally include the following "Physical Device Extended Metrics" types:

• Separately for Physical Device read and write I/O operations and as observed by the hIOmon I/O Monitor:
  -  Accumulated number of I/O operations
  - Accumulated number of "Fast" I/O operations; that is, the number of Physical Device I/O operations observed by the hIOmon I/O Monitor that were successfully completed in less than one millisecond.
  -  Accumulated number of Split I/O operations
  -  Accumulated  total (in seconds) of the time durations associated with the I/O operations
  -  Maximum data transfer size (in bytes) associated with an observed I/O operation
  -  Accumulated number of I/O operations that represent random access (i.e., the random access I/O count).  An I/O operation is considered to be a random access when the data transferred at its starting address is not contiguous to the address of the last data transferred by the preceding I/O operation.
  -  Accumulated total (in bytes) of data transferred by I/O operations that performed a random access (i.e., the random access data transferred amount)
  -  Accumulated number of I/O operations that represent sequential access (i.e., the sequential access I/O count).  An I/O operation is considered to be a sequential access when the data transferred at its starting address is contiguous to the address of the last data transferred by the preceding I/O operation.
  -  Accumulated total (in bytes) of data transferred by I/O operations that performed a sequential access (i.e., the sequential access data transferred amount)

• The number of Physical Device I/O operations (whether Read or Write) that represented a random access to the Physical Device. A Physical Device I/O operation is considered to be a random access when the data transferred at its starting address is not contiguous to the address of the last data transferred by the preceding Physical Device I/O operation as observed by the hIOmon I/O Monitor.

• The number of Physical Device I/O operations (whether Read or Write) that represent a sequential access to the Physical Device. A Physical Device I/O operation is considered to be a sequential access when the data transferred at its starting address is contiguous to the address of the last data transferred by the preceding Physical Device I/O operation as observed by the hIOmon I/O Monitor.

Please see "Appendix A - Export Data Definitions" in the hIOmon User Guide document for additional information about the metrics above.

Also note that the script automatically configures the hIOmon Manager Export File such that it contains only “Device Summary” metrics (that is, summary metrics for the individual files monitored are not exported; only the “Device Summary” metrics, which represent the aggregate for all of the monitored I/O operations for a particular physical device or all monitored files associated with a particular Logical Disk, are written to the export file).

Back to Top

hyperI/O^sm -  improving the performance of storage I/O in computer servers