Plan and Deploy Thin Provisioning

Article
08/31/2016

Applies To: Windows Server 2012 R2, Windows Server 2012

Before you use the thin provisioning feature in Windows Server 2012, you should determine if thin provisioning is appropriate for your environment, and if it is, plan for its use. As a system administrator, you should work closely with your storage administrator and other IT staff to ensure that you make accurate predictions about disk capacity needs, that you agree on threshold notification settings that give you enough time to react to disk space issues and to procure new disks if necessary, and that the system administrator and the storage administrator develop a coordinated response to events and alerts.

In this topic

Note

To fully support the thin provisioning feature in Windows Server 2012, including space reclamation (also known as file-level trim), your storage array must comply with the certification requirements for Windows Server 2012.

Determine if thin provisioning is appropriate for your environment

Before you use thin provisioning, ensure that it is an appropriate storage solution for your environment. Thin provisioning is not designed for all storage usage models. Consider using thin provisioning if the following conditions are true:

There is predictable storage consumption on a storage volume or a low-volatile usage pattern.
The storage volume can tolerate a brief outage. For example, do not use thin provisioning if a mission critical cluster disk cannot tolerate any downtime or delays that are caused by thin provisioning. For example, there may be delays that are associated with temporary resource exhaustion or with space reclamation that occurs after large files are deleted.
There is an adequate storage monitoring processes that detects when critical thresholds are crossed. You should have well-defined policies for monitoring and for response.
You understand the time requirement to acquire new storage. If you do not understand the storage resource procurement process, thin-provisioned logical units (LUNs) may be exposed to the risk of permanent resource exhaustion.

Perform capacity forecasting

Predictable storage consumption is extremely important for thin-provisioned LUN deployment. A storage array that can support thin provisioning enables storage administrators to create large size LUNs with less physical disk space. A thin-provisioned LUN acts like a virtual storage device. Physical disk resources are allocated only when they are needed. This design reduces the amount of unused physical disk capacity and enhances the efficiency of storage use. However, this also introduces the risk that you will run out of physical disk space unless you carefully plan for and predict capacity requirements.

There are two basic forecasting approaches: quantitative and qualitative. For storage resource forecasting, it is a best practice to use both methods.

Quantitative forecasting. Quantitative forecasting uses statistical analysis to analyze historical resource consumption. This provides a basis to determine future resource needs. Quantitative analysis provides a predicted growth rate for storage resource consumption, and it provides some insight into seasonal peaks that may occur, such as yearly application rollouts or semiannual migrations to data warehouses. The main drawback of using only quantitative forecasting is that history is not always an accurate predictor of future needs.
Qualitative forecasting. Qualitative forecasting uses the expertise of the IT staff and other stakeholders to forecast needs based on their real-world knowledge and experience. We recommend that you work as a team, leveraging the insight of the system administrators, storage administrators, line-of-business (LOB) application owners, and senior managers who can help provide a more accurate forecast.

We recommend that you do all the following as part of capacity planning:

Gather forecasting data from the IT department. When you do, communicate the value of forecasting to senior IT managers and to the business units.
Develop accurate metrics. It is crucial to develop the appropriate forecasting metrics so that they are meaningful to all interested parties. Meaningful metrics will help to solidify the value of the initiative for stakeholders such as senior managers.
Discuss capacity-on-demand with your storage vendors. A capacity-on-demand buying process is a low-risk way to build flexibility into a resource plan.

Use the storage resource forecasting results to create a plan to deploy thin-provisioned LUNs and to procure physical storage. If storage capacity prediction is not available or reliable, you should not use the thin provisioning feature.

Determine threshold notification and resource exhaustion settings

A storage administrator can configure threshold notification and resource exhausting settings.

Threshold notification settings

A thin provisioning threshold notification is an alert setting that warns system administrators and storage administrators when a predetermined capacity threshold is reached. There are two types of thresholds: LUN threshold and pool threshold. A storage administrator must configure these threshold values by using the vendor’s storage management tools. When a threshold is reached, the Windows Server operating system logs an event to the System log in Event Viewer to alert the system administrator that the resource use or availability threshold was reached for the LUN or pool.

As a system administrator, you should work with your storage administrator to confirm the threshold notification settings. As a best practice, we recommend the following:

When a system administrator requests thin-provisioned LUNs from the storage administrator, the system administrator should provide the desired threshold notification settings.
If a system administrator does not provide the desired threshold notification settings, the storage administrator should determine a rational threshold setting and inform the system administrator.

Resource exhaustion settings

There are two kinds of resource exhaustion for a thin-provisioned LUN: temporary and permanent.

Temporary resource exhaustion setting. Temporary resource exhaustion occurs when a LUN reaches the configured size limit, but automatic growth (or auto-grow) is enabled. When auto-grow or over-subscription is supported for a LUN, the storage array can use the temporary resource exhaustion setting to interact with the host server while it allocates more resources to the LUN. If temporary resource exhaustion occurs, the application that is using the disk is not interrupted or stopped.
Permanent resource exhaustion setting. Permanent resource exhaustion represents a hard limit on the LUN’s capacity consumption. This prevents the LUN from unexpectedly using capacity when other LUNs need it. If resource exhaustion is permanent, the storage array cannot allocate more resources to the LUN. The Windows partition manager sets the LUN to offline when the permanent resource exhaustion threshold limit is reached.

Note

It is important that the system administrator and the storage administer confirm the permanent resource exhaustion setting.

Monitor and respond to events

After you deploy thin-provisioned LUNs to application and data servers, and threshold notification and resource exhaustion settings are configured by the storage administrator. The system administrator and the storage administrator should monitor and respond to related events.

You can monitor thin provisioning threshold notifications from the System log in Event Viewer or from the vendor’s storage management tools. The following table lists the threshold notification warning events.

Event ID	General Information
144	Threshold notification without additional information
145	Threshold notification without specific information
146	Threshold notification with used LUN capacity and available LUN capacity information
147	Threshold notification with used LUN capacity and available pool capacity information
148	Threshold notification with used pool capacity and available LUN capacity information
149	Threshold notification with used pool capacity and available pool capacity information

When a threshold notification event is detected, the system administrator should immediately notify the storage administrator that an event has occurred and provide the event details. Working together, you should do the following:

Review the System log information, LUN use status, and pool consumption level.
Project LUN and pool capacity requirements.
Expand the LUN size or increase the pool size as needed.
In the storage management tools, reset the threshold notification setting.

How to respond to a permanent resource exhaustion event

If a permanent resource exhaustion threshold is reached, the following occurs:

Event 150 is logged in the System log.
The thin-provisioned LUN is set to offline, and is therefore unavailable.

If this occurs, you can use the following steps to recover the offline LUN:

The system administrator reviews the information that is associated with event ID 150 in the System log.
The system administrator informs the storage administrator about the issue.
The storage administrator either extends the LUN capacity or adds more physical disk resources to resolve the permanent resource exhaustion.
The storage administrator notifies the system administrator about the change in LUN status.
The system administrator uses File and Storage Services in Server Manager or the Disk Management tool to rescan the disk, bring the disk online, and extend the disk size.

Consider space reclamation and potential performance impact

In Windows Server 2012, when a large file is deleted, the file system performs space reclamation according to the storage array setting. Large file deletions can potentially affect the performance of the regular I/O. To protect against this, you can do either of the following:

Configure a registry setting to disable the file-delete notification that triggers real-time space reclamation. Realize that if you configure this setting, this disables the file-delete notification for all LUNs that are assigned to the host.
If you want to use real-time space reclamation, but you have a thin-provisioned LUN that is used by a highly volatile application where large file deletions are common, we recommend that you convert the thin-provisioned LUN to a thick (or full) provisioned LUN. A storage administrator can do this by using the vendor’s storage management tools.

To disable the file-delete notification (trim) in the registry, follow these steps:

Note

The following registry change impacts all NTFS volumes and does not require a restart.

Warning

Incorrectly editing the registry may severely damage your system. Before you make changes to the registry, back up valued data on the computer.

Start Registry Editor.
Locate the following registry subkey:

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\FileSystem
Double-click DisableDeleteNotification.
In the Value data box, enter a value of 1, and then click OK.

If you have concerns or questions regarding the status of space reclamation, we recommend that you contact your storage administrator or your storage array vendor for more detailed information or for troubleshooting recommendations.