Another day, another Spectre fix slowdown: What to expect if you heart ZFS
Blogger records 7-8% toll on read IOPS
Posted in Storage, 17th January 2018 14:35 GMT
The widely used ZFS file system software is slowed down in both read IOPS and throughput by Intel CPU microcode fixes for the Spectre processor design flaws, one set of numbers suggests.
Systems engineer Jon Kensy blogged about the results of his ZFS testing, based on a VM running ZFS on a Ubuntu 16.04 LTS system on a Dell R620.
He set up a VMware IO Analyzer in a VM on the NFS datastore on the ZFS storage and checked the maximum read and write IOPS and read and write throughput (MB/sec), with a 120-second run of each test before the microcode was patched and then after the microcode was fixed.
Here's what he found:
|Read Throughput||Write Throughput||Read IOPS||Write IOPS|
|Pre minus post number||55||-2||3,391||353|
|Per cent change||-8.2%||+0.6%||-7.67%||-2.52%|
The hit was 7-8 per cent for read IOPS and throughput, a negligible gain for write throughput and slight hit for write IOPS.
It was not a heavily stressed system. He writes: "For production environments where clients demand the best performance possible, a 7.6-8 per cent impact is pretty rough especially since the current tenor of this whole ordeal is one of being cheated out of performance."
He said he thinks ZFS systems should be patched unless "a storage node is extremely well-secured, has no VMs running on it, and only presents storage to white-listed hosts then maybe it could go without the microcode update".
Nexenta's software uses ZFS and its recommendation is: "In the cases where our software is run as a VM or a Docker container, we do recommend that customers patch the underlying OS and hypervisors."
Users of Lustre have found a 10-45 per cent performance hit after applying the Spectre and Meltdown patches. ®