Why is it that the read operations are 0 but the read bandwidth is >0?
What is iostat
[not] accounting for? Is it the metadata reads? (Is it possible to
determine what kind of metadata
reads these are?
I plan to have 3 disks and am debating what I should do with them, if I
should do a
raidz (single or double parity) or just a mirror.
As per some of the blog entries, I''ve been reading that raidz may not
be
that suitable for lot of
random reads.
With the # of reads below, I don''t see any reason why I should consider
that. I would like to
proceed with doing a raidz with double parity, please give me some feedback.
Thanks,
Anil
capacity operations bandwidth
pool used avail read write read write
---------- ----- ----- ----- ----- ----- -----
data1 41.6G 5.67G 2 19 52.3K 198K
data2 58.2G 9.83G 3 44 60.5K 180K
---------- ----- ----- ----- ----- ----- -----
data1 41.6G 5.66G 0 21 11.3K 151K
data2 58.2G 9.83G 0 44 158 140K
---------- ----- ----- ----- ----- ----- -----
data1 41.6G 5.66G 0 18 436 117K
data2 58.2G 9.83G 0 44 203 149K
---------- ----- ----- ----- ----- ----- -----
data1 41.6G 5.66G 0 20 1.49K 167K
data2 58.2G 9.83G 0 44 331 154K
---------- ----- ----- ----- ----- ----- -----
data1 41.6G 5.66G 0 21 791 166K
data2 58.2G 9.83G 0 46 199 167K
---------- ----- ----- ----- ----- ----- -----
data1 41.6G 5.66G 0 25 686 364K
data2 58.2G 9.83G 0 45 35.9K 152K
---------- ----- ----- ----- ----- ----- -----
data1 41.6G 5.66G 0 19 698 129K
data2 58.2G 9.83G 0 43 81 146K
---------- ----- ----- ----- ----- ----- -----
data1 41.6G 5.66G 0 19 1.45K 141K
data2 58.2G 9.82G 0 44 59 139K
---------- ----- ----- ----- ----- ----- -----
data1 41.6G 5.66G 0 19 436 124K
data2 58.2G 9.82G 0 43 71 145K
---------- ----- ----- ----- ----- ----- -----
data1 41.6G 5.66G 0 21 412 150K
data2 58.2G 9.82G 0 41 114 138K
---------- ----- ----- ----- ----- ----- -----
data1 41.6G 5.66G 0 20 1.35K 128K
data2 58.2G 9.82G 0 47 918 160K
---------- ----- ----- ----- ----- ----- -----
On Wed, Jun 18, 2008 at 07:37:38PM -0700, Anil Jangity wrote:> Why is it that the read operations are 0 but the read bandwidth is >0? > What is iostat > [not] accounting for? Is it the metadata reads? (Is it possible to > determine what kind of metadata > reads these are?That question I''ll leave for others to answer.> I plan to have 3 disks and am debating what I should do with them, if I > should do a > raidz (single or double parity) or just a mirror. > > With the # of reads below, I don''t see any reason why I should consider > that. I would like to > proceed with doing a raidz with double parity, please give me some feedback.A three-way mirror and three disks in a double parity array are going to get you the same usable space. They are going to get you the same level of redundancy. The only difference is that the RAIDZ2 is going to consume a lot more CPU cycles calculating parity for no good cause. In this case, a three-way mirror is the way to go. -brian -- "Coding in C is like sending a 3 year old to do groceries. You gotta tell them exactly what you want or you''ll end up with a cupboard full of pop tarts and pancake mix." -- IRC User (http://www.bash.org/?841435)
G''Day Anil, On Wed, Jun 18, 2008 at 07:37:38PM -0700, Anil Jangity wrote:> Why is it that the read operations are 0 but the read bandwidth is >0? > What is iostat > [not] accounting for? Is it the metadata reads? (Is it possible to > determine what kind of metadata > reads these are?This could just be rounding - what interval did you use for zpool? I believe 1 read per 5 seconds will be shown as ''0'' (rounded from 0.2); the read bandwidth in your output is very small, suggesting that something like this may be happening. Brendan> > I plan to have 3 disks and am debating what I should do with them, if I > should do a > raidz (single or double parity) or just a mirror. > > As per some of the blog entries, I''ve been reading that raidz may not be > that suitable for lot of > random reads. > > With the # of reads below, I don''t see any reason why I should consider > that. I would like to > proceed with doing a raidz with double parity, please give me some feedback. > > Thanks, > Anil > > capacity operations bandwidth > pool used avail read write read write > ---------- ----- ----- ----- ----- ----- ----- > data1 41.6G 5.67G 2 19 52.3K 198K > data2 58.2G 9.83G 3 44 60.5K 180K > ---------- ----- ----- ----- ----- ----- ----- > data1 41.6G 5.66G 0 21 11.3K 151K > data2 58.2G 9.83G 0 44 158 140K > ---------- ----- ----- ----- ----- ----- ----- > data1 41.6G 5.66G 0 18 436 117K > data2 58.2G 9.83G 0 44 203 149K > ---------- ----- ----- ----- ----- ----- ----- > data1 41.6G 5.66G 0 20 1.49K 167K > data2 58.2G 9.83G 0 44 331 154K > ---------- ----- ----- ----- ----- ----- ----- > data1 41.6G 5.66G 0 21 791 166K > data2 58.2G 9.83G 0 46 199 167K > ---------- ----- ----- ----- ----- ----- ----- > data1 41.6G 5.66G 0 25 686 364K > data2 58.2G 9.83G 0 45 35.9K 152K > ---------- ----- ----- ----- ----- ----- ----- > data1 41.6G 5.66G 0 19 698 129K > data2 58.2G 9.83G 0 43 81 146K > ---------- ----- ----- ----- ----- ----- ----- > data1 41.6G 5.66G 0 19 1.45K 141K > data2 58.2G 9.82G 0 44 59 139K > ---------- ----- ----- ----- ----- ----- ----- > data1 41.6G 5.66G 0 19 436 124K > data2 58.2G 9.82G 0 43 71 145K > ---------- ----- ----- ----- ----- ----- ----- > data1 41.6G 5.66G 0 21 412 150K > data2 58.2G 9.82G 0 41 114 138K > ---------- ----- ----- ----- ----- ----- ----- > data1 41.6G 5.66G 0 20 1.35K 128K > data2 58.2G 9.82G 0 47 918 160K > ---------- ----- ----- ----- ----- ----- ----- > > > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss-- Brendan [CA, USA]
> > A three-way mirror and three disks in a double parity array are going to get you > the same usable space. They are going to get you the same level of redundancy. > The only difference is that the RAIDZ2 is going to consume a lot more CPU cycles > calculating parity for no good cause. > > In this case, a three-way mirror is the way to go. >Sorry, I meant to say that I will have 3 disks if I am doing raidz. If I am doing mirror, I will just have 2. Thanks for clarification on something else though, about double parity with 3 disks resulting in only 1 usable disk blocks. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20080618/1d3f1fc3/attachment.html>
I was using a 5 minute interval. I did another test with 1 second interval: data1 41.6G 5.65G 0 0 63.4K 0 data2 58.2G 9.81G 0 447 0 2.31M So, the 63K read bandwidth doesn''t show any read operations still. Is that still rounding? What exactly is an operation? (just any IO access?)> This could just be rounding - what interval did you use for zpool? I believe > 1 read per 5 seconds will be shown as ''0'' (rounded from 0.2); the read > bandwidth in your output is very small, suggesting that something like > this may be happening. > > Brendan > > >> I plan to have 3 disks and am debating what I should do with them, if I >> should do a >> raidz (single or double parity) or just a mirror. >> >> As per some of the blog entries, I''ve been reading that raidz may not be >> that suitable for lot of >> random reads. >> >> With the # of reads below, I don''t see any reason why I should consider >> that. I would like to >> proceed with doing a raidz with double parity, please give me some feedback. >> >> Thanks, >> Anil >> >> capacity operations bandwidth >> pool used avail read write read write >> ---------- ----- ----- ----- ----- ----- ----- >> data1 41.6G 5.67G 2 19 52.3K 198K >> data2 58.2G 9.83G 3 44 60.5K 180K >> ---------- ----- ----- ----- ----- ----- ----- >> data1 41.6G 5.66G 0 21 11.3K 151K >> data2 58.2G 9.83G 0 44 158 140K >> ---------- ----- ----- ----- ----- ----- ----- >> data1 41.6G 5.66G 0 18 436 117K >> data2 58.2G 9.83G 0 44 203 149K >> ---------- ----- ----- ----- ----- ----- ----- >> data1 41.6G 5.66G 0 20 1.49K 167K >> data2 58.2G 9.83G 0 44 331 154K >> ---------- ----- ----- ----- ----- ----- ----- >> data1 41.6G 5.66G 0 21 791 166K >> data2 58.2G 9.83G 0 46 199 167K >> ---------- ----- ----- ----- ----- ----- ----- >> data1 41.6G 5.66G 0 25 686 364K >> data2 58.2G 9.83G 0 45 35.9K 152K >> ---------- ----- ----- ----- ----- ----- ----- >> data1 41.6G 5.66G 0 19 698 129K >> data2 58.2G 9.83G 0 43 81 146K >> ---------- ----- ----- ----- ----- ----- ----- >> data1 41.6G 5.66G 0 19 1.45K 141K >> data2 58.2G 9.82G 0 44 59 139K >> ---------- ----- ----- ----- ----- ----- ----- >> data1 41.6G 5.66G 0 19 436 124K >> data2 58.2G 9.82G 0 43 71 145K >> ---------- ----- ----- ----- ----- ----- ----- >> data1 41.6G 5.66G 0 21 412 150K >> data2 58.2G 9.82G 0 41 114 138K >> ---------- ----- ----- ----- ----- ----- ----- >> data1 41.6G 5.66G 0 20 1.35K 128K >> data2 58.2G 9.82G 0 47 918 160K >> ---------- ----- ----- ----- ----- ----- ----- >> >> >> _______________________________________________ >> zfs-discuss mailing list >> zfs-discuss at opensolaris.org >> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss >> > >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20080618/f6134c0c/attachment.html>
Hello Brian,
Thursday, June 19, 2008, 3:44:01 AM, you wrote:
BH> A three-way mirror and three disks in a double parity array are going to
get you
BH> the same usable space. They are going to get you the same level of
redundancy.
BH> The only difference is that the RAIDZ2 is going to consume a lot more CPU
cycles
BH> calculating parity for no good cause.
And you will also get higher IOPS with 3-way mirror.
--
Best regards,
Robert mailto:milek at task.gda.pl
http://milek.blogspot.com
On Wed, 18 Jun 2008, Anil Jangity wrote:> I plan to have 3 disks and am debating what I should do with them, if I > should do a > raidz (single or double parity) or just a mirror. > > With the # of reads below, I don''t see any reason why I should consider > that. I would like to > proceed with doing a raidz with double parity, please give me some feedback.I can''t imagine any reason to use raidz with double parity with just three disks. This reduces performance and traps these disks forever. Using normal or triple mirror provides more performance and allows you to liberate one or two of these disks for some other purpose later. Bob =====================================Bob Friesenhahn bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/ GraphicsMagick Maintainer, http://www.GraphicsMagick.org/
On Thu, Jun 19, 2008 at 10:06:19AM +0100, Robert Milkowski wrote:> Hello Brian, > > BH> A three-way mirror and three disks in a double parity array are going to get you > BH> the same usable space. They are going to get you the same level of redundancy. > BH> The only difference is that the RAIDZ2 is going to consume a lot more CPU cycles > BH> calculating parity for no good cause. > > And you will also get higher IOPS with 3-way mirror.That''s a good point that I completely forgot to make, thanks! -brian -- "Coding in C is like sending a 3 year old to do groceries. You gotta tell them exactly what you want or you''ll end up with a cupboard full of pop tarts and pancake mix." -- IRC User (http://www.bash.org/?841435)