Using the following commands believe I've created two striped meta
volumes:

metainit d1 1 2 c2t1d0s0 c2t2d0s0
metainit d2 1 4 c2t3d0s0 c2t4d0s0 c2t5d0s0 c2t6d0s0

metastat
d2: Concat/Stripe
Size: 136737528 blocks (65 GB)
Stripe 0: (interlace: 32 blocks)
Device Start Block Dbase Reloc
c2t3d0s0 0 No Yes
c2t4d0s0 4712 No Yes
c2t5d0s0 4712 No Yes
c2t6d0s0 4712 No Yes

d1: Concat/Stripe
Size: 68366408 blocks (32 GB)
Stripe 0: (interlace: 32 blocks)
Device Start Block Dbase Reloc
c2t1d0s0 0 No Yes
c2t2d0s0 4712 No Yes

Device Relocation Information:
Device Reloc Device ID
c2t3d0 Yes id1,sd@SFUJITSU_MAN3184M_SUN18G_02Z40460____
c2t4d0 Yes id1,sd@SFUJITSU_MAN3184M_SUN18G_02Z40985____
c2t5d0 Yes id1,sd@SFUJITSU_MAN3184M_SUN18G_02Z40095____
c2t6d0 Yes id1,sd@SFUJITSU_MAN3184M_SUN18G_00Z03659____
c2t1d0 Yes id1,sd@SFUJITSU_MAN3184M_SUN18G_01Z05467____
c2t2d0 Yes id1,sd@SFUJITSU_MAN3184M_SUN18G_02Z28132____

newfs /dev/md/rdsk/d1
newfs /dev/md/rdsk/d2

mount /dev/md/dsk/d1 /array/disk1
mount /dev/md/dsk/d2 /array/disk2

df -h .
Filesystem size used avail capacity Mounted on
/dev/md/dsk/d1 32G 33M 32G 1% /array/disk1

df -h .
Filesystem size used avail capacity Mounted on
/dev/md/dsk/d2 64G 64M 64G 1% /array/disk2


Question 1

I was under the impression that one of the advantages of using
a stripe was the increase in disk access performance.

If this is true should I notice a significant difference between
these two volumes?

time mkfile 128m /array/disk1/128m

real 0m4.514s
user 0m0.034s
sys 0m0.879s

time mkfile 128m /array/disk2/128m

real 0m4.058s
user 0m0.033s
sys 0m0.993s

If d2 has 4 physical drives and d1 has 2 why do they both seem
to write data at the same speed?


Question 2

d2: Concat/Stripe
Size: 136737528 blocks (65 GB)
Stripe 0: (interlace: 32 blocks)
Device Start Block Dbase Reloc
c2t3d0s0 0 No Yes
c2t4d0s0 4712 No Yes
c2t5d0s0 4712 No Yes
c2t6d0s0 4712 No Yes

What does the Start Block field represent?


Many thanks for helping a meta neophyte

> time mkfile 128m /array/disk1/128m
>
> real 0m4.514s
> user 0m0.034s
> sys 0m0.879s
>
> time mkfile 128m /array/disk2/128m
>
> real 0m4.058s
> user 0m0.033s
> sys 0m0.993s
>

> If d2 has 4 physical drives and d1 has 2 why do they both seem
> to write data at the same speed?
>

The problem here seems to be that all your disks are on the same scsi
adapter (C2) so the bottleneck seems to be the throughput of its bus
(128mb / 4s = 30mb/s). Even if all your disks can do 160mb/s if your
scsi bus can only do 30mb/s you won't see any benefit with striping.
Try 2 disks on different adapters to see if it improve.

Well for Question 2 i was about to ask the same ...

F.

sgsnewsgrp@gmail.com wrote:
> Question 1
>
> I was under the impression that one of the advantages of using
> a stripe was the increase in disk access performance.

The disks are old, so I suspect the SCSI controller they are attached to
as old too. Maybe the SCSI bus is limited to 40MB/s.

With a small stripe size the overhead gets larger on the SCSI bus.

Try creating the devices with a larger interlace factor (512..2048)

metainit d1 1 4 .... -i 512

I don't know at which chunks mkfile writes the file. So you should better
use dd to control the block size.

And you should write a larger file to get better results:

dd if=/dev/zero of=/array/disk1/1g bs=1024k count=1024


> Question 2
>
> d2: Concat/Stripe
> Size: 136737528 blocks (65 GB)
> Stripe 0: (interlace: 32 blocks)
> Device Start Block Dbase Reloc
> c2t3d0s0 0 No Yes
> c2t4d0s0 4712 No Yes
> c2t5d0s0 4712 No Yes
> c2t6d0s0 4712 No Yes
>
> What does the Start Block field represent?

Probably on c2t3d0 slice 0 starts at cylinder 0. This cylinder gets skipped
so the VTOC doesn't get overwritten by accident.

On the other disks slice 0 doesn't start at cylinder 0, so this is not
necessary.

--
Daniel

> The problem here seems to be that all your disks are on the same scsi
> adapter (C2) so the bottleneck seems to be the throughput of its bus
> (128mb / 4s = 30mb/s). Even if all your disks can do 160mb/s if your
> scsi bus can only do 30mb/s you won't see any benefit with striping.
> Try 2 disks on different adapters to see if it improve.


This seems to be the case as the bus can only handle a max of 40mb/sec
according to the hardware specs



> The disks are old, so I suspect the SCSI controller they are attached to
> as old too. Maybe the SCSI bus is limited to 40MB/s.


The sun blade 2000 external scsi bus supports at most 40mb/sec

This now makes sense to me



>
> With a small stripe size the overhead gets larger on the SCSI bus.
>
> Try creating the devices with a larger interlace factor (512..2048)
>
> metainit d1 1 4 .... -i 512
>
> I don't know at which chunks mkfile writes the file. So you should better
> use dd to control the block size.
>
> And you should write a larger file to get better results:
>
> dd if=/dev/zero of=/array/disk1/1g bs=1024k count=1024

Same results



> Probably on c2t3d0 slice 0 starts at cylinder 0. This cylinder gets skipped
> so the VTOC doesn't get overwritten by accident.
>
> On the other disks slice 0 doesn't start at cylinder 0, so this is not
> necessary.



When I setup these disks I used format to layout the slices on the
first disk
then ran the following:

prtvtoc /dev/rdsk/c2t1d0s0 | fmthard -s - /dev/rdsk/c2t2d0s0
prtvtoc /dev/rdsk/c2t1d0s0 | fmthard -s - /dev/rdsk/c2t3d0s0
prtvtoc /dev/rdsk/c2t1d0s0 | fmthard -s - /dev/rdsk/c2t4d0s0
prtvtoc /dev/rdsk/c2t1d0s0 | fmthard -s - /dev/rdsk/c2t5d0s0
prtvtoc /dev/rdsk/c2t1d0s0 | fmthard -s - /dev/rdsk/c2t6d0s0


They seem to be identical as far as the slices are concerned:



Total disk cylinders available: 7506 + 2 (reserved cylinders)

Part Tag Flag Cylinders Size Blocks
0 unassigned wm 0 - 7504 16.86GB (7505/0/0)
35363560
1 unassigned wm 0 0 (0/0/0)
0
2 backup wu 0 - 7505 16.86GB (7506/0/0)
35368272
3 unassigned wu 0 0 (0/0/0)
0
4 unassigned wu 0 0 (0/0/0)
0
5 unassigned wu 0 0 (0/0/0)
0
6 unassigned wu 0 0 (0/0/0)
0
7 unassigned wu 0 0 (0/0/0)
0

Part Tag Flag Cylinders Size Blocks
0 unassigned wm 0 - 7504 16.86GB (7505/0/0)
35363560
1 unassigned wu 0 0 (0/0/0)
0
2 backup wu 0 - 7505 16.86GB (7506/0/0)
35368272
3 unassigned wu 0 0 (0/0/0)
0
4 unassigned wu 0 0 (0/0/0)
0
5 unassigned wu 0 0 (0/0/0)
0
6 unassigned wu 0 0 (0/0/0)
0
7 unassigned wu 0 0 (0/0/0)
0