"James Cassell" <fedoraproject@xxxxxxxxxxxxx> writes: > On Tue, 23 Dec 2008 10:01:28 -0500, Marc Schwartz > <marc_schwartz@xxxxxxxxxxx> wrote: > >> >> I have a separate /boot partition outside the LVM, since that cannot be >> encrypted. >> Using hdparm to test sequential reads on the encrypted and unencrypted >> partitions, I get 30 MB/Sec on the former and 36 MB/Sec on the latter. >> So I am looking at a 15-20% hit on throughput and that has been pretty >> consistent over several releases. > > Could the performance difference here be due to the partitions being > on different parts of the disk? Throughput is higher on the outside > of the disk (which is the logical beginning of the disk.) I don't > think you have done a fair benchmark. Fair point and it is possible, but under prior releases (pre-F9), when I had to do a manual config and I did not have all partitions other than /boot encrypted, I got pretty similar results in throughput changes across the partitions. There was a period of time when I did not have '/' encrypted, but did have /home, swap, /tmp and /var encrypted as separate partitions without using LVM. Clearly a better comparison would compare the same partitions in an unencrypted and encrypted configuration. Without going through some config gyrations that's easier said than done. I don't think that it is unreasonable to expect some level of performance hit, albeit with multi-core systems, that should be lessened and perhaps those who do not experience a notable reduction in throughput are using such systems. I'll have one next year... :-) Thanks James, Marc -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
