LUN size consideration while creating on new EMC Clariion Storage.

[Narendra Uttekar]

Hi,
We are going to replace our old EMC storage Clariion CX500 to new EMC storage Clariion CX4-240. We have OS HP-UX 11i V1. And oracle database 10.2g and SAP ECC6.0, currently on our prodcution server we have an 16 LUN's size of 64GB each for oracle database.
I want to know what will be the best LUN size to be considered while creating on new EMC storage, And aslo we are going to migrate the Oracle Database to SAP MAXDB database once we moved to new storage. I want to know whether it will be good just i create 500GB of 2 LUN's on new storage for database spread across 5 disks? And also i want to know while allocating LUN's from storage we have to follow any guidelines like to allocate LUN's of siliar size or different size of LUN's can be allocated to HP-UX host?

Thanks,
Narendra

[Kapil Jha]

I suppose its good idea to make 100GB of LUNS, do not make LUNs of huge size(although does not affect the performace significantly) but its good to have Luns of normal size rather than having huge LUN.

There is nothing like you have to allocate LUNs of same size.

BR,
Kapil+

[smatador]

Hi,
In the past, we used 33GB or less, after that we used 50GB. Because the storage increase every year, I suppose that 100GB or little more is a good option now.
If you create only 2 Lun of 500, the system see only 2 disks. So I don't think it's good for admin task [ 1 vg or 2 vg max ], for performance even if the database is spread across a raid5 in fact the system just see 2 disks.
You have to debate with Oracle and Sap admin about spread the database on multiples "axes", maybe you need some little lun for redo files and big lun for dbf.
HTH

[WayneHP]

You want dedicated spindles to Oracle Datafiles, and seperate spindles for oracle logs and format the logs filesystem with 1K blks. WHoever set up the 16x64GB LUNS to be used for SAP/ORACLER chose wisely.

Something like this

/oracle/stage/ vgecpapp lv_ecpstage 6,291,456
/oracle/ECP vgecpapp lv_ecpECD 33,554,432
/oracle/ECP/saparch vgecpapp lv_ecpsaparch 33,554,432
/oracle/ECP/sapreorg vgecpapp lv_ecpsapreorg 10,485,760
/usr/sap/ECP vgecpapp lv_ecpusrsap 3,145,728
/sapmnt/ECP vgecpapp lv_ecpsapmnt 87 4,194,304

/oracle/ECP/origlogA vgecplog lv_ecporiglogA 4,194,304
/oracle/ECP/origlogB vgecplog lv_ecporiglogB 4,194,304
/oracle/ECP/mirrlogA vgecplog lv_ecpmirrlogA 4,194,304
/oracle/ECP/mirrlogB vgecplog lv_ecpmirrlogB 4,194,304

/oracle/ECP/sapdata1 vgecpdata lv_ecpsapdata1 134,217,728
/oracle/ECP/sapdata2 vgecpdata lv_ecpsapdata2 134,217,728
/oracle/ECP/sapdata3 vgecpdata lv_ecpsapdata3 134,217,728
/oracle/ECP/sapdata4 vgecpdata lv_ecpsapdata4 134,217,728

/oracle vgapp lv_oracle 8,388,608
/usr/sap vgapp lv_usrsap 8,388,608
/usr/sap/trans vgapp lv_usesaptrans 8,388,608
/sapmnt vgapp lv_sapmnt 8,388,608

You may want 8 or more sapdata mount points and a biger log size it depends on what you already have.

We booted from SAN
so the VGs in this example need (GB)

vgecdapp 128
vgecdlog 16
vgecddata 512
vgapp 32
vg00 64
vgswap 64

and I always request Luns in POWERERS of 2
so 32GB, 64GB, 128GB etc.

for vgecpdata I would prefer to have 8x64GB LUNS in this example.

Do not mush this all on to a few luns in a big VG.

[WayneHP]

and mount anything under Oracle SGA control like this in /etc/fstab

/dev/vgecpdata/lv_ecpsapdata1 /oracle/ECP/sapdata1 vxfs rw,largefiles,convosync=direct,mincache=direct,delaylog,nodatainlog 0 2
/dev/vgecpdata/lv_ecpsapdata2 /oracle/ECP/sapdata2 vxfs rw,largefiles,convosync=direct,mincache=direct,delaylog,nodatainlog 0 2
/dev/vgecpdata/lv_ecpsapdata3 /oracle/ECP/sapdata3 vxfs rw,largefiles,convosync=direct,mincache=direct,delaylog,nodatainlog 0 2
/dev/vgecpdata/lv_ecpsapdata4 /oracle/ECP/sapdata4 vxfs rw,largefiles,convosync=direct,mincache=direct,delaylog,nodatainlog 0 2

/dev/vgecplog/lv_ecporiglogA /oracle/ECP/origlogA vxfs rw,largefiles,convosync=direct,mincache=direct,delaylog,nodatainlog 0 2
/dev/vgecplog/lv_ecporiglogB /oracle/ECP/origlogB vxfs rw,largefiles,convosync=direct,mincache=direct,delaylog,nodatainlog 0 2
/dev/vgecplog/lv_ecpmirrlogA /oracle/ECP/mirrlogA vxfs rw,largefiles,convosync=direct,mincache=direct,delaylog,nodatainlog 0 2
/dev/vgecplog/lv_ecpmirrlogB /oracle/ECP/mirrlogB vxfs rw,largefiles,convosync=direct,mincache=direct,delaylog,nodatainlog 0 2

and the rest:

/dev/vgecpapp/lv_ecpsapmnt /sapmnt/ECP vxfs rw,largefiles,delaylog 0 2