Anyone have any experience with Sanbolic Clustered Filesystems?
I have a customer that reported poor performance on a VMAX LUN. A trace was run and was determined that the host is doing 260K and 516K IOs to the VMAX and is causing a lot of disk crossings.
Is this normal for a Sanbolic File System?
Although your inquiry did not include details pertaining to the architecture, based on the information provided, it sounds as though the shared volume created with Sanbolic's Melio cluster file system may have been configured using a stripe size of 64KB, which is the default size. To ensure proper alignment with data blocks, it is suggested that the stripe size be set to 512KB, which should result in a significant improvement in performance.
It is a Windows 2K8 R2 VM on ESX 5.0.
The customer created a standard NTFS volume and we are still seeing the same IO profile so it is not related to the Sanbolic driver. Is there anything that can be tuned at the ESX layer that could be causing this?
I would recommend contacting VMware technical support to see if there are any settings that may need to be changed within the properties of the ESX host itself in order to improve performance.