Storage tiering and cache management for Windows Server storage spaces

Storage Space Storage tiering

before we talk about storage tiering, we need to understand how many layers are included in the Windows Server storage space and how to define each hierarchy. Based on the relative value of the speed of reading and writing performance, storage space is divided into two data storage tiers, a storage layer is a fast layer, usually provided by SSD hard disk, used to store hot data, that is, frequently accessed data; the other storage layer is a slow layer, usually provided by HDD hard drives, for storing cold data, or infrequently accessed data. By default, storage spaces have virtual disks of storage tiers with different storage rates, and storage tiering is designed to balance SSD performance and hard disk drive (HDD) capacity. When there is free space in the storage tier, new data is usually written to the slow layer, and the storage space moves sub-file-level data transparently between two tiers based on how often the data is accessed. Storage tiers can meet the performance requirements of thermal data by moving the most common hot data to SSD storage, and move cold data to the HDD layer to meet the capacity requirements of ordinary data. Storage tiering works as shown in 1.          

650) this.width=650; "title=" 301-figure 1.jpg "style=" Height:347px;width:500px;float:none; "alt=" Wkiol1jpl4uzl2ftaav8g9hdnjo674.jpg-wh_50 "src=" https://s2.51cto.com/wyfs02/M01/8E/F1/                    Wkiol1jpl4uzl2ftaav8g9hdnjo674.jpg-wh_500x0-wm_3-wmp_4-s_2477059006.jpg "width=" height= "611"/>

Figure 1 storage space tiering Technology

Storage controller server operating system Windows Server 2012 and Windows 8 and later builds built-in storage tiered management Service (Storage Tiers Management Service) that is installed on Windows The server file server role is activated when it is running. The storage tiered Management service automatically calculates data heat and starts hot and cold data movement at 1 o'clock every night by default policy. To observe the storage tiering tasks and behaviors, you can log in to the storage controller server, open the server's control panel to find the scheduled task program, and in the scheduled task program, expand "Scheduled Tasks"-"Microsoft"-"Windows" to find the name "Storage Tiersmanagement "project, open the Storage Tiers optimization property, in the Triggers tab we can see that the storage tiering Management Service triggers a scheduled task of hot and cold data migration at 1 o'clock in the morning every day, as shown in 2. This default setting is not immutable, and you can modify the time to trigger layered tasks as required and actual.

650) this.width=650; "title=" 301-figure 2.jpg "style=" Float:none; "alt=" Wkiol1jpl4yqeo-saablffj_khe674.jpg-wh_50 "src=" Https://s5.51cto.com/wyfs02/M01/8E/F1/wKioL1jPl4yQEO-sAABlFfj_khE674.jpg-wh_500x0-wm_3-wmp_4-s_1171586843.jpg " />

Figure 2 Automatic storage space tiering tasks

In addition to modifying the trigger time for a layered operation task, you can also complete the storage hierarchy without relying on the scheduled task program. In order to find out how the storage Tiers Management service implements the storage tiering, we can find the commands and parameters related to the layered operation from the pre-set scheduled task. Click on the Actions tab of the Storage Tiers Optimization Properties window to see the specific tasks and related parameters that storage Tiers Management service performs in the details bar, in fact the storage hierarchy uses the Defrage command and the-C -h-g-#这4个参数完成, 3 shown.

650) this.width=650; "title=" 301-figure 3.jpg "style=" Float:none; "alt=" Wkiom1jpl46hkwkuaabmmri_vsi814.jpg-wh_50 "src=" Https://s3.51cto.com/wyfs02/M02/8E/F3/wKiom1jPl46hKwKUAABMMRI_vsI814.jpg-wh_500x0-wm_3-wmp_4-s_3876302051.jpg " />

Figure 3 Storage-Space tiered scheduling operations

        gets to the tiered storage commands and parameters, We can manually trigger the hot and cold data migration using the command, just enter C + + at the command prompt Windows\system32\defrag.exe-c-h-g-#即可.

< Span style= "font-size:14px;font-family: ' The song Body '; Color:rgb (0,0,0);" If you are familiar with windows before 8, Microsoft did not introduce storage space, there is no storage tiering technology, so use Defrag.exe/? Command gets help after Windows The system will have more -g this parameter. This parameter is important, meaning that the storage layer is optimized to find the appropriate storage layer for the data according to the data heat.

It should be emphasized that storage tiering is based on the granularity of the data rather than the granularity of the file, and the storage tiering service analyzes and moves the data in 1MB units. If the file is larger than 1MB, and only some "hot" data, then a file only a certain proportion of data will be moved to the SSD layer, for example, there is a virtual machine on the disk above the data only 20% hot data, only 20% of the data will be moved to the SSD partition, Instead of moving the entire virtual disk file to the SSD partition. To give a practical example, a virtual machine of the VHDX size of a total of 100GB, the installation of the SQL Server instance, under the instance 20GB database and log files for the trading system, the trading system has generated thousands of transactions per day, then this 20GB data will be moved to the SSD partition, The remaining 80GB system files and other file data will not necessarily be moved to the SSD partition.As for how the storage space is calculated for data heat, please refer to another blog, "Windows Serversoftware-defined storage-Storage SpaceIOthe distribution ofhttp://ichbinleo.blog.51cto.com/11948851/1900880。

Storage tiering is the process of automating data migrations in the background, and if certain virtual machines need to run continuously on a high-speed SSD layer, or a virtual machine with low performance requirements needs to be fixed on a large HDD layer to provide additional fast SSD tiers for other VMs, you might consider pinning the file to a disk layer using a fixed command. The difference between pinning and layering is that hierarchies are data-level operations, and pinning is file-level operations. For example, to be \\FileServer\Share\VM01. The VHDX is fixed on the SSD layer and is completed as follows:

( 1) Run the PS c:\>get-storagetier|ft friendlyname PowerShell command to get an existing hierarchical friendly name.

(2) return the following results, Select a layer of type SSD based on mediatype .

FriendlyName mediatype---------------------ssdtier SSD hddtier HDD

(3) Run the following example command to pin the VM01 virtual disk to the SSD layer, using clear-filestoragetier unpin A file to a layer .

PS c:\> Set-filestoragetier-filepath "\\FileServer\Share\VM01. VHDX "-desiredstoragetierfriendlyname" Ssdtier "SSD

The command that runs the pinned file, the store does not immediately move the file data to the specified hierarchy, but waits for the next disk tiering optimization run task to begin, and if it cannot wait, follow the methods mentioned above, at the scheduled task program, command prompt, or PowerShell manually trigger the data movement operation.

storage space Cache management

write-back ) and through write ( write-through windows Server SDS

Write-Back cache

applications are usually randomly written to storage as small data, asSSDhard Disk key properties, write-back cache can cache small random writes toSSDhard drive avoids excessive disk writesI/O. FromWindows Server R2start, storage space supports creating a write-back cache that uses the poolSSDto buffer small-scale random writes and reduce the impact on other parallel data writes, and to improve the write performance of the storage by reducing the delay of random writes. A common workload random write operation is directed to theSSDcache, then write laterHDD. The write-back cache is also used for power-down protection of the data. Storage space if presentSSDhard disk, the virtual disk will be created by default with the size1GBwrite-back cache space,4is shown.

The write-back cache is transparent to the administrator, with a sufficient amount of storage space SSD , the new virtual disk will create a write-back cache. The disk layout and number of SSD requirements are described below:

• a simple layout requires a piece of SSD hard drives;

• dual-mirror and single parity-parity requirements with two blocks SSD hard drives;

• triple mirroring and dual parity require three blocks SSD hard drive.

650) this.width=650; "title=" 301-figure 4.jpg "style=" Float:none; "alt=" Wkiom1jpl47iklefaaavarhma8g190.jpg-wh_50 "src=" Https://s1.51cto.com/wyfs02/M02/8E/F3/wKiom1jPl47iKLEfAAAVArhMa8g190.jpg-wh_500x0-wm_3-wmp_4-s_687745770.jpg "/ >

Figure 4 write-Back cache

Journal

when we set the hard disk's useUsagethere's an argument.Journalvalues,Journal(no Chinese translation of Xinda) and write-back cache have a similar role, are responsible for handling the re-organization of the activities on the pathI/O, but in the storage space, the roles and settings of the two are clearly different. Journalstorage space for parity layouts only, write-back caching is available for all storage space layouts and can support tiered storage. We can set up a dedicated storage space for parityJournalhard drives, usually using a fastSSDhard disk, configured with a dedicatedJournalthe hard disk's parity storage space can greatly improve the memory write performance, especially when sequential write, in addition, the use of more than two blocks ofJournalthe hard drive can achieve its own image guaranteeI/Oof redundancy.

This article from "Huang Lijun Blog" blog, reproduced please contact the author!

Storage tiering and cache management for Windows Server storage spaces