EXT4 is ~2X as slow as XFS (593MB/s vs 304MB/s) for writes?
in [Kernel]
|
Prev: [PATCH 08/11] ocfs2: Pass the locking protocol into ocfs2_cluster_connect().
Next: [PATCH tip/core/rcu 1/3] rcu: fixes for accelerated grace periods for last non-dynticked CPU
From: Eric Sandeen on 1 Mar 2010 19:10 Justin Piszcz wrote: > Hello, > > Is it possible to 'optimize' ext4 so it is as fast as XFS for writes? > I see about half the performance as XFS for sequential writes. > > I have checked the doc and tried several options, a few of which are shown > below (I have also tried the commit/journal_async/etc options but none > of them get the write speeds anywhere near XFS)? > > Sure 'dd' is not a real benchmark, etc, etc, but with 10Gbps between 2 > hosts I get 550MiB/s+ on reads from EXT4 but only 100-200MiB/s write. > > When it was XFS I used to get 400-600MiB/s for writes for the same RAID > volume. > > How do I 'speed' up ext4? Is it possible? > FWIW I'm seeing similar things on fast storage (Fusion IO), though this is under 2.6.31. 500MB/s+ for xfs, 300 for ext4. Overwriting an existing file is no faster. I don't think this driver is blktraceable but I'll try a newer driver that should be I think. (xfs's overwrite went from 534 to 597 mb/s; ext4 sat at 320-ish) direct IO was good for both xfs & ext4 at around 530mb/s I'll see if I can get this running on a more recent kernel to do further investigation. -Eric -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Eric Sandeen on 1 Mar 2010 19:40
Eric Sandeen wrote: > Justin Piszcz wrote: >> Hello, >> >> Is it possible to 'optimize' ext4 so it is as fast as XFS for writes? >> I see about half the performance as XFS for sequential writes. >> >> I have checked the doc and tried several options, a few of which are shown >> below (I have also tried the commit/journal_async/etc options but none >> of them get the write speeds anywhere near XFS)? >> >> Sure 'dd' is not a real benchmark, etc, etc, but with 10Gbps between 2 >> hosts I get 550MiB/s+ on reads from EXT4 but only 100-200MiB/s write. >> >> When it was XFS I used to get 400-600MiB/s for writes for the same RAID >> volume. >> >> How do I 'speed' up ext4? Is it possible? >> > > FWIW I'm seeing similar things on fast storage (Fusion IO), > though this is under 2.6.31. 500MB/s+ for xfs, 300 for ext4. > > Overwriting an existing file is no faster. I don't think this > driver is blktraceable but I'll try a newer driver that should be I think. FWIW, blktrace (I'm still on 2.6.31) is enlightening: Total (xfs): Reads Queued: 4, 16KiB Writes Queued: 122,567, 10,485MiB Read Dispatches: 4, 16KiB Write Dispatches: 83,219, 10,485MiB Reads Requeued: 0 Writes Requeued: 0 Reads Completed: 4, 16KiB Writes Completed: 83,219, 10,485MiB Read Merges: 0, 0KiB Write Merges: 39,348, 314,804KiB IO unplugs: 344 Timer unplugs: 338 Total (ext4): Reads Queued: 14, 56KiB Writes Queued: 2,621K, 10,486MiB Read Dispatches: 14, 56KiB Write Dispatches: 107,944, 10,486MiB Reads Requeued: 0 Writes Requeued: 0 Reads Completed: 14, 56KiB Writes Completed: 107,944, 10,486MiB Read Merges: 0, 0KiB Write Merges: 2,513K, 10,054MiB IO unplugs: 2,461 Timer unplugs: 2,020 See "Writes Queued" See also submit_bio() calls in xfs. ext4 doing things a block at a time is certainly giving the elevator a workout... I'd tend to chalk it up to that at first glance. -Eric -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ |