Messages in this thread | | | Date | Thu, 15 Jan 2004 09:33:05 +0100 | From | "Dragan Krnic" <> | Subject | Re: extreme performance degradation - ext3/sata |
| |
Hi,
I've encountered a problem with sata drives and ext3/reiser, which I documented in the correspondence with Neil Brown below. He referred me to your forum. Since I'm not subscribed please cc my return address.
Regards Dragan
>> I have extreme performance degradation when writing to >> ext3fs on SATA drives, about 7 times slower than reading >> from the same volume - 17 MB/s writing, 120 MB/s reading. >> >> The problem also affects reiserfs in the same way, however >> xfs appears to be immune to this problem. It reads AND writes >> to the same volume at about 100 MB/s, which is the same >> speed at which null-bytes from /dev/zero can be poured >> into the raw volume. >> >> My setup is: SuSE 8.2 (2.4.20), Siemens Scenic W600 (i865g), >> P4/3GHz, 2GB PC2100 RAM, a SCSI boot/swap/root disk, >> 6 Maxtor 250 GB SATA drives, 2 on on-board connectors, >> 4 on 2 Promise SATA 150 Tx2+ PCI 32/33 controllers >> (pdc-ultra.o). >> >> The volumes have been formatted with commands like: >> >> mke2fs -J device=/dev/sda2 /dev/md0 >> mkreiserfs -j /dev/sda2 /dev/md0 >> mkfs.xfs -l logdev=/dev/sda2 /dev/md0 >> >> whereby "/dev/md0" may also be "/dev/vg01/lvol1". >> >> The performance degradation, as well as the steady performance >> of xfs, is totaly independent of the disk clustering mode, >> e.g. 5-way RAID5, 6-way RAID5, 3-way RAID0 of 2-way RAID1, >> 2-way RAID0 (for higher priority on-board controller), >> 4-way RAID0 (for slightly slower PCI controllers), using >> either software raid or LVM. >> >> I wish I could give you more info to help you pin-point the >> performance killer. Please let me know, what more I can do >> to help. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |