NickF
Guru
- Joined
- Jun 12, 2014
- Messages
- 763
Hi all,
General inquiry. I have two pools, one which has 2 SATA SSDs Mirrored and one with 2 NVME SSD Mirrored. They are both served via iSCSI over a single 10g link to an ESXI host on a standard vSwitch. Both pools are set with a 64K record size. Things work great, and I have no issues. I am just observing performance and I noticed something which I find odd. The FreeNAS host has 32GB of RAM, specs are in my signature.
Running a benchmark on the C: drive in a Windows guest on the SATA pool, there are very few competing resources running on this pool during the benchmark:
Running the same benchmark on the C: drive in a Windows guest on the NVME pool, results in similar behavior, the write speeds are substantially faster than the reads:
(There are like 9 VMs running that use this pool as their OS drive so that is why relative performance is lower than the SATA pool, which is used for other purposes)
Is there something inherent to ZFS or Mirrored VDEVs that would explain why my writes are consistently and substantially faster than my reads? Just trying to understand I guess.
General inquiry. I have two pools, one which has 2 SATA SSDs Mirrored and one with 2 NVME SSD Mirrored. They are both served via iSCSI over a single 10g link to an ESXI host on a standard vSwitch. Both pools are set with a 64K record size. Things work great, and I have no issues. I am just observing performance and I noticed something which I find odd. The FreeNAS host has 32GB of RAM, specs are in my signature.
Running a benchmark on the C: drive in a Windows guest on the SATA pool, there are very few competing resources running on this pool during the benchmark:
Running the same benchmark on the C: drive in a Windows guest on the NVME pool, results in similar behavior, the write speeds are substantially faster than the reads:
(There are like 9 VMs running that use this pool as their OS drive so that is why relative performance is lower than the SATA pool, which is used for other purposes)
Is there something inherent to ZFS or Mirrored VDEVs that would explain why my writes are consistently and substantially faster than my reads? Just trying to understand I guess.