Hello, everyone!
Do you know how to solve the write bandwidth is twice higher than read bandwidth during file copy between LUNs? Don't worry, the post will share with you!
Fault Description
When copying files before two LUNs are added to an active-active pair, it is found that the read bandwidth and write bandwidth of the two LUNs are inconsistent.
Symptom
Multiple LUNs of an OceanStor V3 storage system are mapped to an ESXi 5.5 host with separate datastores created. Files are copied between datastores of two LUNs (from LUN 0 to LUN 1).
On DeviceManager, it is found that the write bandwidth of LUN 1 is twice higher than the read bandwidth of LUN 0.
Theoretically, the write bandwidth should be the same as the read bandwidth during file copy in a storage system.
Alarm Information
N/A
Cause
The disk formatting mode for VMware virtualization is set to lazy zeroed. Before service I/Os are written to unformatted space, all 0 data is written (zeroed).
Identification Method
The following three virtual disk formats are available for VMware virtualization:
Thick Provision Lazy Zeroed: Required space is allocated immediately after a datastore is created. Data in the space is not erased temporarily, but is zeroed out later when I/Os are delivered.
Thick Provision Eager Zeroed: Data in the allocated space is zeroed out immediately after a datastore is created.
Thin Provision: Space is allocated on demand, and data is zeroed out after space is allocated.
The default mode is Thick Provision Lazy Zeroed. In this mode, free space will be initialized (zeroed) when it needs to be used. Therefore, data in the free space must be zeroed out before service data is written. The volume of written data is twice more than that of read data, so it is found that the write bandwidth is twice higher than the read bandwidth.
Solution
Start a test again after formatting.
Application Scope
All versions of OceanStor V3 series.
This is my solution, how about yours? Go ahead and share it with us!
