Question is basically in the title.
I'm trying to figure out the best way to optimise my setup for iSCSI performance. Assume a single client, mostly read (assume 60% sequential/40% random) with occasional writes and a mixture of tiny files and large files (40GB+). I have 8 x 7200rpm HDDs in mirrors, along with 2 x Optane 900p 280GB. The system has 128GB 1866mhz DDR3, and a Xeon E5 1660v2 CPU.
With the Optanes, I was planning to use them as mirrored SLOGs, and then partition them further for testing and use them as L2ARC and/or metadata vdevs. In the future, I'll add additional Optanes to run as either striped L2ARC or a striped mirror metadata pool - depending on what has the biggest performance impact.
So yeah... has anyone tried metadata vdevs with iSCSI? Does it help? Or would it be better to just have L2ARC? Thank you in advance for any advice :)
I'm trying to figure out the best way to optimise my setup for iSCSI performance. Assume a single client, mostly read (assume 60% sequential/40% random) with occasional writes and a mixture of tiny files and large files (40GB+). I have 8 x 7200rpm HDDs in mirrors, along with 2 x Optane 900p 280GB. The system has 128GB 1866mhz DDR3, and a Xeon E5 1660v2 CPU.
With the Optanes, I was planning to use them as mirrored SLOGs, and then partition them further for testing and use them as L2ARC and/or metadata vdevs. In the future, I'll add additional Optanes to run as either striped L2ARC or a striped mirror metadata pool - depending on what has the biggest performance impact.
So yeah... has anyone tried metadata vdevs with iSCSI? Does it help? Or would it be better to just have L2ARC? Thank you in advance for any advice :)