How RAID10 IS CONFIGURED IN SYMMETRIX DMX4?

Its also called Striped CKD. Its Raid 1 devices striped across multiple disk through meta. You can create it through command line:

sample cmd:

===========

symconfigure -sid xxx -cmd "create dev count=xx, size=xxx cyl, emulation=CKD-3390, config=2-way-mir, mvs_ssid=xxx, disk_group=xxx, attribute=ckd_meta;" preview

Hi saurabh,

thanks for the reply and command.So,now if I configure 2-way-mirror in my storage environment which has 1TB free space

in a disk group,am i going to get 500Gb or less than that.?

A RAID1 volume is only on two disks.  You can create a meta volume by putting several together on FBA.  On CKD (mainframe) it has to be 4 volumes put together (8 disks).

The largest volume you can create on DMX is about 64GB in size.  You could put those together to make a larger volume on FBA.

Hi Srikanth,

- The RAID 10 is configured as shown in the above pic.
- A RAID-1 LUN is created that spans across two physical disks. (The first two disks in the above pic).
- Expand the existing RAID-1 disks using Meta Expansion concept.
- While expanding the RAID-1 disks, use Striped Meta to expand.
- Now, the first data gets committed to the first disk (disk with number 1)
- This data gets copied to the disk 2, as disk 2 is the other mirror of the RAID-1 configuration.
- The next incoming data, moves over to Disk 3, as we have expanded the Meta using the "Striped" option.
- In effect, we see that the data gets mirrored first and then striped, which is how RAID 10 is acheived.
- Thus, even if we do not have the RAID 10 LUNs configured initially, we can still 'manipulate' RAID 1 LUNs to act as RAID 10 LUNs.

FYI - While Meta creation, if concatenated option is selected against the Striped option, the data will be written to the disk1 first, this data mirrored to disk 2, this cycle repeats until the disk 1 and eventually disk 2 are completely written to, and then the data gets written to disk 3, which is parallelly mirrored to the disk 4.

Are you referring to open systems or mainframe storage? 

As people have already stated you can create Raid10 for open systems by having Raid-1 luns configured as a striped meta volume.  On the mainframe side there is an option for raid10 CKD type disks where the CKD volume is sliced across four sets of disk drives, essentially a meta volume for the mainframe.

Hi RobertDudley,

Thanks for reply.I am referring to open system not to main frame.

Yeah you can certainly do that, by making sure that you create "Concatenated Meta" rather than a striped meta., because when you create a concatenated meta, the disk 1 and disk 3 in the above pic, will be logically merged into a single larger disk, and the same is true with disk 2 and disk 4. So, you have a RAID-1 Meta, with (Disk 1 + Disk 3) being the first mirror of the RAID-1 and (Disk 2 + Disk 4) being the second identical mirror.

Thanks,

SreeHari

Hi sreehari,

   But you mentioned earlier if we use concatenated option while creating meta the data will be written to the disk1 first, this data mirrored to disk 2, this cycle repeats until the disk 1 and eventually disk 2 are completely written to, and then the data gets written to disk 3, which is parallelly mirrored to the disk 4.So,I feel here our purpose of RAID 10 is not acheived because we can observe no striping of data is done here.Data is going into disk 1 first(mirrored to 2),when it gets full,then it is going to disk3(mirrored to 4),which is different from striping,where data chunks go to both disk 1 and 3 and gets mirrored to 2 and 4 respectively. So,I feel even for 2-way-mirror hypers we have to go for striped option.

Hi Srikanth,

Lets say you have a RAID-1 LUN of 100GB, and want to expand it to 150 GB. So you need to effectively add 50 GB Storage to each of the RAID-1 disks. That means, now logically each disk is 150 GB from the host perspective. So, we are no more striping the data, but acheiving RAID-1 !!!!

A concatenated meta volume would not be considered a Raid10 device.  Raid10 is striping data across raid1 protected devices.  A concatenated meta volume fills the first disk before moving on to the other members in order.