Hi!
I've stumbled over this article about Windows 2022 Storage Bus Cache feature and wanted to give the simple space a try. So I took my testbed and attached some SATA storage devices and installed the module and the failover clustering feature.
The next step was to assign devices 1 and 2 to the storage bus.
Then I created the new pool and verified the correct usage of the devices.
Finally I created a new simple volume with ReFS on the pool and assigned a drive letter.
Now the problem began. It doesn't matter what file I/O operation I used (local/shared file copy, robocopy /J, disk benchmarks, mixed patterns) the performance of the pool didn't differ then if I would use the hdd directly. Sometimes even worse.
I've tried different file system settings and played with the Storage Bus Cache settings (shared, cache, reserved bytes/%) and updated it after each change but it had nearly no effects on the I/O result.
During the tests I often executed Get-StorageBusBinding and observed the DirtyByteCount value which always was moving arround a few hundred KB only on write operations. Honestly I don't know the meaning of this value exactly but I would expect this are the data that is on the cache device and not written to the HDD yet and therefore should increase rapidly because of the huge different write speed of both devices.
Has anyone else tested this feature and could share his/her experience?
I've stumbled over this article about Windows 2022 Storage Bus Cache feature and wanted to give the simple space a try. So I took my testbed and attached some SATA storage devices and installed the module and the failover clustering feature.
Code:
> Get-PhysicalDisk
Number FriendlyName SerialNumber MediaType CanPool OperationalStatus HealthStatus Usage Size
------ ------------ ------------ --------- ------- ----------------- ------------ ----- ----
1 ST9160314AS 6VCYG2J9 HDD True OK Healthy Auto-Select 149.05 GB
2 Samsung SSD 840 Series S19HNSAD432819Y SSD True OK Healthy Auto-Select 111.79 GB
0 SanDisk SDSSDH2128G 130665401751 SSD True OK Healthy Auto-Select 119.24 GB
Code:
> Enable-StorageBusDisk -Number 1
> Enable-StorageBusDisk -Number 2
> Get-PhysicalDisk
Number FriendlyName SerialNumber MediaType CanPool OperationalStatus HealthStatus Usage Size
------ ------------ ------------ --------- ------- ----------------- ------------ ----- ----
501 ST9160314AS 6VCYG2J9 HDD True OK Healthy Auto-Select 149.05 GB
502 Samsung SSD 840 Series S19HNSAD432819Y SSD True OK Healthy Auto-Select 111.79 GB
0 SanDisk SDSSDH2128G 130665401751 SSD True OK Healthy Auto-Select 119.24 GB
Code:
> Update-StorageBusCache
> Get-StoragePool
FriendlyName OperationalStatus HealthStatus IsPrimordial IsReadOnly Size AllocatedSize
------------ ----------------- ------------ ------------ ---------- ---- -------------
Primordial OK Healthy True False 380.08 GB 170.8 GB
Storage Bus Cache on WIN-EEU9AC1KEEK OK Healthy False False 169.8 GB 512 MB
> Get-StorageBusBinding
DeviceGuid DeviceNumber CacheDeviceGuid CacheDeviceNumber CacheMode DirtyByteCount TotalByteCount
---------- ------------ --------------- ----------------- --------- -------------- --------------
{393170cb-2cca-a5c3-78a4-67aa495fd464} 501 {42c1a6e5-8921-24fd-681d-ce1e3b08d788} 502 ReadWrite 0 B 0 B
Now the problem began. It doesn't matter what file I/O operation I used (local/shared file copy, robocopy /J, disk benchmarks, mixed patterns) the performance of the pool didn't differ then if I would use the hdd directly. Sometimes even worse.
I've tried different file system settings and played with the Storage Bus Cache settings (shared, cache, reserved bytes/%) and updated it after each change but it had nearly no effects on the I/O result.
During the tests I often executed Get-StorageBusBinding and observed the DirtyByteCount value which always was moving arround a few hundred KB only on write operations. Honestly I don't know the meaning of this value exactly but I would expect this are the data that is on the cache device and not written to the HDD yet and therefore should increase rapidly because of the huge different write speed of both devices.
Has anyone else tested this feature and could share his/her experience?
Last edited: