HPE Storage Users Group

Hi,

I am slightly confused as to why x2 VVols (1 full prov the other Thin ) both 2GB each in size and both RAID1, use different number of chuncklets when viewed in grid layout. To be precise the Fully Prov VVol uses 4 chuncklets but the thin prov uses over 60? Does 3 par do some kind of sub-block allocation with thin prov VVols ?

My CPG growth increment is 32GB. Note both VVols are in the same CPG.

There is a follow up question and it relates to performance. If I have a 10GB Ful Prov VVol in a CPC which has say 120 disks. There is a very good chance that this data will only be spread on a small number of spindles meaning performance will not be as optimal as possible.


Thanks

Don't quite understand what you're seeing without the cli output, but if you think about it the FPVV is fully allocated up front and using raid 1 it will only consume a maximum on 4 chunklets which are fixed and non shared.
On the TPVV it won't fully allocate space until data gets written and remember the space (in terms of chunklets) can be shared as data is written in pages. As such it may be that what you're seeing is actually the CPG min growth increment which will in the real world will actually be shared by multiple VV's overlapping the same chunklets.

Yes that's correct. if you do a search there's a min VV formula you can use to touch all disks.

Hi, I made sure both VVs were populated with the same data so that both have real content within. If you look at the screenshots you will see that the TPVV is at 100% and the mounted volume has 2GB of data within.

Looking slightly lower at in the Grid view you will see that the fullVV had 4 chunklets and the TPVV has many. So are you saying chunklets can be shared between VVs? Does this mean out of the 2 VVs above, the TPVV could potentially out perform the FullVV?

Thanks

I'm unsure sure what you're seeing from the grid as there's nothing to put it in context, I'd look at the underlying LD's that back the VV's to see what's going on under the covers. You could start with the following:-

showcpg -sdg (have a look at the CPG setup and growth)
showvv - s (volume space allocation)
showld -d (provide a list of LD's on the system)
showld -vv "vvname" (show the LD's mapped to the affected VV's)
showldch "ldname" (have a look at the chunklets mapped to the above LD's)

Here is the output from the last command in your list
Does this mean the FPVV is less performant ?

------------------For FPVV--------------------------------
3par-01cli% showldch Test001.usr.0
Ldch Row Set PdPos Pdid Pdch State Usage Media Sp From To
0 0 0 4:1:0 97 5 normal ld valid N --- ---
1 0 0 1:21:0 45 16 normal ld valid N --- ---
3par-01cli% showldch Test001.usr.1
Ldch Row Set PdPos Pdid Pdch State Usage Media Sp From To
0 0 0 4:18:0 114 5 normal ld valid N --- ---
1 0 0 0:22:0 22 26 normal ld valid N --- ---

------------------For TPVV--------------------------------
3par-01 cli% showldch tp-0-sa-0.0
Ldch Row Set PdPos Pdid Pdch State Usage Media Sp From To
0 0 0 4:21:0 117 0 normal ld valid N --- ---
1 0 0 0:19:0 19 4 normal ld valid N --- ---
2 0 0 2:19:0 67 4 normal ld valid N --- ---
3 0 1 4:13:0 109 4 normal ld valid N --- ---
4 0 1 1:3:0 27 3 normal ld valid N --- ---
5 0 1 3:3:0 75 3 normal ld valid N --- ---
6 0 2 4:15:0 111 4 normal ld valid N --- ---
7 0 2 0:11:0 11 3 normal ld valid N --- ---
8 0 2 1:11:0 35 4 normal ld valid N --- ---
9 0 3 4:23:0 119 4 normal ld valid N --- ---
10 0 3 3:11:0 83 4 normal ld valid N --- ---
11 0 3 1:19:0 43 4 normal ld valid N --- ---
3par-01 cli% showldch tp-0-sa-0.1
Ldch Row Set PdPos Pdid Pdch State Usage Media Sp From To
0 0 0 4:12:0 108 0 normal ld valid N --- ---
1 0 0 0:20:0 20 2 normal ld valid N --- ---
2 0 0 2:20:0 68 2 normal ld valid N --- ---
3 0 1 4:20:0 116 4 normal ld valid N --- ---
4 0 1 3:20:0 92 2 normal ld valid N --- ---
5 0 1 1:20:0 44 2 normal ld valid N --- ---
6 0 2 4:14:0 110 4 normal ld valid N --- ---
7 0 2 0:12:0 12 3 normal ld valid N --- ---
8 0 2 2:12:0 60 3 normal ld valid N --- ---
9 0 3 4:22:0 118 4 normal ld valid N --- ---
10 0 3 1:12:0 36 3 normal ld valid N --- ---
11 0 3 3:12:0 84 3 normal ld valid N --- ---
3par-01 cli% showldch tp-0-sd-0.0
Ldch Row Set PdPos Pdid Pdch State Usage Media Sp From To
0 0 0 4:5:0 101 0 normal ld valid N --- ---
1 0 0 2:1:0 49 17 normal ld valid N --- ---
2 0 1 4:21:0 117 4 normal ld valid N --- ---
3 0 1 1:1:0 25 4 normal ld valid N --- ---
4 0 2 4:13:0 109 5 normal ld valid N --- ---
5 0 2 3:1:0 73 4 normal ld valid N --- ---
6 0 3 4:9:0 105 0 normal ld valid N --- ---
7 0 3 0:9:0 9 4 normal ld valid N --- ---
8 0 4 4:17:0 113 0 normal ld valid N --- ---
9 0 4 2:9:0 57 4 normal ld valid N --- ---
10 0 5 4:1:0 97 0 normal ld valid N --- ---
11 0 5 0:15:0 15 4 normal ld valid N --- ---
12 0 6 4:7:0 103 4 normal ld valid N --- ---
13 0 6 3:19:0 91 4 normal ld valid N --- ---
14 0 7 4:15:0 111 5 normal ld valid N --- ---
15 0 7 2:15:0 63 4 normal ld valid N --- ---
16 0 8 4:23:0 119 5 normal ld valid N --- ---
17 0 8 1:13:0 37 3 normal ld valid N --- ---
18 0 9 4:3:0 99 4 normal ld valid N --- ---
19 0 9 1:17:0 41 18 normal ld valid N --- ---
20 0 10 4:11:0 107 4 normal ld valid N --- ---
21 0 10 3:9:0 81 5 normal ld valid N --- ---
22 0 11 4:19:0 115 4 normal ld valid N --- ---
23 0 11 0:21:0 21 3 normal ld valid N --- ---
24 0 12 1:5:0 29 15 normal ld valid N --- ---
25 0 12 3:5:0 77 2 normal ld valid N --- ---
26 0 13 2:21:0 69 3 normal ld valid N --- ---
27 0 13 0:5:0 5 2 normal ld valid N --- ---
28 0 14 3:13:0 85 24 normal ld valid N --- ---
29 0 14 0:13:0 13 3 normal ld valid N --- ---
30 0 15 1:21:0 45 3 normal ld valid N --- ---
31 0 15 2:5:0 53 2 normal ld valid N --- ---
3par-01 cli% showldch tp-0-sd-0.1
Ldch Row Set PdPos Pdid Pdch State Usage Media Sp From To
0 0 0 4:0:0 96 0 normal ld valid N --- ---
1 0 0 1:0:0 24 4 normal ld valid N --- ---
2 0 1 4:8:0 104 0 normal ld valid N --- ---
3 0 1 2:8:0 56 4 normal ld valid N --- ---
4 0 2 4:16:0 112 0 normal ld valid N --- ---
5 0 2 0:8:0 8 4 normal ld valid N --- ---
6 0 3 4:4:0 100 0 normal ld valid N --- ---
7 0 3 3:0:0 72 4 normal ld valid N --- ---
8 0 4 4:12:0 108 4 normal ld valid N --- ---
9 0 4 0:20:0 20 15 normal ld valid N --- ---
10 0 5 4:20:0 116 5 normal ld valid N --- ---
11 0 5 1:16:0 40 4 normal ld valid N --- ---
12 0 6 4:18:0 114 4 normal ld valid N --- ---
13 0 6 2:20:0 68 15 normal ld valid N --- ---
14 0 7 4:2:0 98 5 normal ld valid N --- ---
15 0 7 0:0:0 0 4 normal ld valid N --- ---
16 0 8 4:10:0 106 5 normal ld valid N --- ---
17 0 8 1:20:0 44 3 normal ld valid N --- ---
18 0 9 4:6:0 102 5 normal ld valid N --- ---
19 0 9 3:16:0 88 4 normal ld valid N --- ---
20 0 10 4:14:0 110 5 normal ld valid N --- ---
21 0 10 2:4:0 52 1 normal ld valid N --- ---
22 0 11 4:22:0 118 5 normal ld valid N --- ---
23 0 11 3:20:0 92 3 normal ld valid N --- ---
24 0 12 2:0:0 48 4 normal ld valid N --- ---
25 0 12 3:8:0 80 5 normal ld valid N --- ---
26 0 13 0:16:0 16 5 normal ld valid N --- ---
27 0 13 1:8:0 32 5 normal ld valid N --- ---
28 0 14 2:16:0 64 5 normal ld valid N --- ---
29 0 14 3:4:0 76 1 normal ld valid N --- ---
30 0 15 1:4:0 28 1 normal ld valid N --- ---
31 0 15 0:4:0 4 1 normal ld valid N --- ---


Thanks

You only posted part of the info (lacking context). So I can see the TPVV's LD's span many more disks so it potentially has more performance, assuming this is raid 1 it will write 256K then move to the next chunklet but I can't see the underlying geometry of the VV's or CPG's so can't see how you ended up with the LD's you got.

Because of the very small size of the FPVV and the fact that it's fully provisioned it will only ever span those 4 chunklets, so yes the TPVV has more potential performance. You can see the odd numbered PD's are aligned with one set of LD's belonging to one node and the even PD's the other LD's and node. Much will depend on how much data actually gets written, but given how small this is(it'll fit in cache) performance between the two will probably be variable/marginal anyway, why don't you test it ?

Have a look here for the VV sizing and CPG growth formulas.
viewtopic.php?f=18&t=633&p=2556&hilit=formula#p2556

Hi,

I'll try and send you the full results soon.

As for stats, not surprising but here are the iops using a default iometer profile.

FPVV
Read: 791
Write: 388

TPVV
Read: 1790
Write: 879

I'm sure I can get way more if I setup the CPGs correctly. Currently have 120 disks but growth is set to default(32GB). Based on the magic formula I should set this to 90. Ill run another test if i have the time.

Thanks