Unsolved
This post is more than 5 years old
13 Posts
0
758
April 17th, 2007 12:00
MetaLUNs / reconfiguring disk array for exchange
After reading the White Paper called: "EMC CLARiiON Storage Solutions Microsoft Exchange 2003 Best Practices", ive come to the conclusion our SAN is configured, in a way that causes disk bottlenecks.
We have a CX500 with 4 Disk Chassis, most of it configured for other applications
First Disk Chassis is 15x146GB FC2 15K disks(the other three chassis arent really relevant as such)
The following is true today:
Disks 0_0_0-0_0_4 is configured for data, the space reserved is 350GB for exchange stores, We currently use around 170-190GB of this for ESGs (Exchange Storage Groups). in a Raid5, Load is around 200-260IOPS
Disks 0_0_5-0_0_13 is configured with 50GBs for log drives, 800GB is configured for an SQL database (100-150IOPS at the highest), Raid5 also. 15th disk is hotspare.
What im puzzling to do is buy 4 x 73GB disks and add them to some available slots i have in Chassis 2 and 3, and make a small Raid 1/0 for log files (around 10Gigs should be enough).
Then i would have around 350GB storage on RaidGroup 1, and another 230GB in RaidGroup 2, and i was hoping it was possible to make a MetaLun Stripe, to gain additional Spindles to my exchange store.
Is it even possible, or are the requirements such that i need the exact same LUN sizes, Or even worse, that i need the same available disk space for all physical disks in the two LUNs?
Are there other ways i could add more spindles to my ESGs without deleting the arrays and restore data into a newly made LUN?
I hope it makes somewhat sense what i want to do. If you have any questions feel free to ask.
Christian
We have a CX500 with 4 Disk Chassis, most of it configured for other applications
First Disk Chassis is 15x146GB FC2 15K disks(the other three chassis arent really relevant as such)
The following is true today:
Disks 0_0_0-0_0_4 is configured for data, the space reserved is 350GB for exchange stores, We currently use around 170-190GB of this for ESGs (Exchange Storage Groups). in a Raid5, Load is around 200-260IOPS
Disks 0_0_5-0_0_13 is configured with 50GBs for log drives, 800GB is configured for an SQL database (100-150IOPS at the highest), Raid5 also. 15th disk is hotspare.
What im puzzling to do is buy 4 x 73GB disks and add them to some available slots i have in Chassis 2 and 3, and make a small Raid 1/0 for log files (around 10Gigs should be enough).
Then i would have around 350GB storage on RaidGroup 1, and another 230GB in RaidGroup 2, and i was hoping it was possible to make a MetaLun Stripe, to gain additional Spindles to my exchange store.
Is it even possible, or are the requirements such that i need the exact same LUN sizes, Or even worse, that i need the same available disk space for all physical disks in the two LUNs?
Are there other ways i could add more spindles to my ESGs without deleting the arrays and restore data into a newly made LUN?
I hope it makes somewhat sense what i want to do. If you have any questions feel free to ask.
Christian
0 events found
No Events found!
Kiran3
410 Posts
0
April 17th, 2007 21:00
Is it even possible, or are the requirements such that i need the exact same LUN sizes, Or even worse, that i need the same available disk space for all physical disks in the two LUNs?
yes, you cant stripe the two luns if they are of different raid type and size.
meta lun is one way. other way is to assign both devices to the host and then use a host level volume manager software to combine the space.
these both ways can be done online
Kiran3
410 Posts
0
April 17th, 2007 21:00
EMC does not recommend to use first 5 disks of first enclosure for high performance access luns. this is because the array uses them for internal storage and under high io loads, it can affect negatively.
Darkingdk78
13 Posts
0
April 17th, 2007 23:00
EMC does not recommend to use first 5 disks of first
enclosure for high performance access luns. this is
because the array uses them for internal storage and
under high io loads, it can affect negatively.
I realise that, its a question of Performance vs cost i believe, but its the main reason i want to spread the load out over alot more spindles. That way the only times i would have full IOPS load on the LUNs would be when we make backup(hopefully)
an_hidden_KB
87 Posts
1
April 18th, 2007 00:00
another lun and attach it to the server (9 disk
Raid5) and move some of the storage groups with the
heavier users to that then.
looking at your load figures i think it is the best way to go. get the 4x73G drives, migrate all transaction logs there and you should have plenty of room for load growth
Darkingdk78
13 Posts
0
April 18th, 2007 00:00
different raid type and size.
Ah darnit, im starting to think im a bit stuck then, i had hoped it would be possible.
ESGs without deleting the arrays and restore data
into a newly made LUN?
meta lun is one way. other way is to assign both
devices to the host and then use a host level volume
manager software to combine the space.
these both ways can be done online
It doesnt seem very appealing to me, id rather make another lun and attach it to the server (9 disk Raid5) and move some of the storage groups with the heavier users to that then. I had hoped to use all 14 disks mainly to provide less stress for the SQL, and in exchange get more speed for the MS-Exchange server.