I have a datastore I want to create from multiple source luns. Each of the luns is from an SSD storage array, using 6G speed, which allos up to 600MB/sec per lun throughput.
When I try to create the datastore, it seems to concatenate the luns together rather than striping them for improved performance.
The application needs to support 2GB/sec, meaning I need to create a stripe accross 8 of these luns in order to achieve that performance, however, the datastore doesn't seem to allow striping.
Is there a way to fix this so that I can get the necessary performance throughput?
vSphere allows you create a datastore from multiple LUNs device, but this will not work like RAID stripes, see misconception #1: VMFS Extents - Are they bad, or simply misunderstood? - VMware vSphere Blog
And I will always avoid to use multiple LUNs per datastore due to increase of management on this... if you want spread the I/O across multiple LUNs, create one datastore per LUN, create a virtual disk in each datastore and present to the virtual machine, and inside the guest OS configure the volume stripe using all virtual disks, in Linux you can use LVM stripe and in Windows use dynamic disk with striped volumes.
vSphere allows you create a datastore from multiple LUNs device, but this will not work like RAID stripes, see misconception #1: VMFS Extents - Are they bad, or simply misunderstood? - VMware vSphere Blog
And I will always avoid to use multiple LUNs per datastore due to increase of management on this... if you want spread the I/O across multiple LUNs, create one datastore per LUN, create a virtual disk in each datastore and present to the virtual machine, and inside the guest OS configure the volume stripe using all virtual disks, in Linux you can use LVM stripe and in Windows use dynamic disk with striped volumes.
That's a good idea, ill go with that route. I can stripe luns created on multiple datastores using LVM in the host (Raid-0).