As mentioned on previous posts, the new TrueNAS is a PowerEdge R320 w/ 48gb of ECC RAM, four HGST 10tb SATA 7200rpm hard drives, an NVMe drive, and a 10gb NIC. The primary pupose of this box is to be a backup target, however it might get used as a shared iSCSI target from time to time, so it seemed worth while to do a bit of benchmarking.
The test client is a Windows 2012r2 Server, running as a VM that lives on VMware ESXi v7, installed on a HP ML350 G8 connected via a 10gb NIC.
For a baseline, here is the OS drive, that lives on three Seagate 6tb 7200rpm SAS drives, in a RAID5 configuration, connected via a HP Smart Array P420 with a 2gb cache.
TrueNAS was setup for iSCSI, a LUN created on the RAID10 SATA drives, connected to VMware, a datastore created, then an additional virtual disk created (thick provisioned, eager zero), and the same test ran.
TrueNAS was again setup with a LUN, but this time on the single PCI-e NVMe drive.
Just for comparison, this is the very same Crucial NVMe drive as a local drive in a VMware server.
Here is a shot of what the 10gb NIC utilization looks like during each of the tests. Interestingly the maximum transfer rate is roughly the same for each test, which would indicated we are reaching maximum network throughput; however the benchmark results would say there is still room for more speed.
FURTHER UPDATES:
Finally got around to swapping the E5-2450 v2 CPU for an E5-2418L v2 CPU; according to Intel's ARK page the maxim CPU wattages is 95w vs 50w. At idle according to the Kill-A-Watt the usages went from 100watts down to 90watts, and according to the iDRAC from 94watts down to 70watts. A quick iSCSI and SMB benchmarking shows no slow down in file transfers; not that I expected them to as I have file compression and dedupe turned off; so the CPU doesn't have much to do.
No comments:
Post a Comment