Bottom line – if you want to use disks with different tiers in the same pool effectively you should get a FAST VP license

Otherwise create pools with just one tier or traditional raidgroups and manually balance using online LUN migration

Hi Guys,

Thanks for your reply.

Can you please explain how i can assess the impact of "Thin" LUN when compared to a "Thick" LUN? How much of a performance impact will there be?

Is there a specific best practice that if a certain LUN is larger than a certain size, it is advisable to create it as a Thin Lun or maybe a certain number of IOPs?

Thanks

JS

i have not seen any numbers (maybe our EMC friends and USPEED gurus can chime in), the only numbers i have seen is 10% overhead on pool LUNs versus traditional raid group LUNs (metadata management for pool LUNs). I have not seen any best practices that state thin better for workload xyz and thick better for workload abc. Maybe heavy random write workloads would benefit from thick LUNs because when you create a thick LUN its 1g slices will be not only reserved but also fully allocated (new to flare 32). In my shop we decided to provision everything as thick LUNs because while over subscription is great, we can't procure storage fast enough if one or more systems starts consuming storage really fast. We rely on host side over subscription (thin VMs).

i have not seen any numbers (maybe our EMC friends and USPEED gurus can chime in), the only numbers i have seen is 10% overhead on pool LUNs vs traditional raid group LUNs (metadata management for pool LUNs). I have not seen any best practices that state thin better for workload xyz and thick better for workload abc. Maybe heavy random write workloads would benefit from thick LUNs because when you create a thick LUN its 1g slices will be not only reserved but also fully allocated (new to flare 32). In my shop we decided to provision everything as thick LUNs because while over subscription is great, we can't procure storage fast enough if one or more systems starts consuming storage really fast. We rely on host side over subscription (thin VMs).