This post is more than 5 years old

12 Posts

744

October 19th, 2008 23:00

Question abt DMX RAID control

Hi Experts,

Would just like to get some information abt this issue. I understand that Symmetrix deals with raid groups as opposed to raid disks as in the CX boxes.

My question is, what component in the Symm(hardware) that actually controls the raid for all the disks prior to it being assigned as raid groups.

Am I also correct to say that the raid configurations are stored/saved in the bin file of the Symm's?

6 Operator

 • 

2.8K Posts

October 20th, 2008 04:00

That cleared up quite a bit there.

I'm pretty glad my mud isn't that muddy :D

Is this a good summary from your expalantion? :)

Yep .. it looks a good summary ..
When I say "processor" I mean PowerPC processors running in the heart of your DMX. When I say "microcode" or "code" I mean the operating system that runs on the processors. When I say "binfile" I mean the actual configuration of your very particular storage, used by the code to coordinate work of processors... :D

what actually tells the drive to perform a
rebuild? And what RAID levels are the HD at??


As above, the answer is "the backend", a complex mix of processors, drives, operating system and configuration file.. Or if you prefer, the code and the binfile ;-)

Should a drive fail, the code will detect the broken drive and invoke an hotspare, pulling data from alive drives into cache and then back to the hotspare (or to the replaced drive). While pushing data to the hotspare the box will also open a SR and request a drive replacement.

Please note that the code will periodically poll all devices and detect soon-to-fail drives, thus your box may open a service request (and request a drive replacement).

6 Operator

 • 

2.8K Posts

October 20th, 2008 00:00

No RAID "groups" or "disks" at all .. When you create a single RAID5/6 protected device, the microcode will allocate 4, 8 or 16 slices on 4 8 or 16 different drives (following a lot of strict rules) and use them to protect you data spreading parity and datablocks on all 4, 8 or 16 slices. Thus is legal to have volumes with different RAID protection on the same drive in the backend.

"Raid groups" comes straight from CX (afaik) and in old Symms there was so called "Raid sets". :-)

My question is, what component in the Symm(hardware)
that actually controls the raid for all the disks prior to it
being assigned as raid groups.


Each processor in the backend owns its drives (let's forget about redundancy right now). One of the "rules" I mentioned above forces the code to allocate slices on different processors. Thus your RAID5 device will hopefully use 4, 8 or 16 different processors in the backend (I'm simplifying a lot here). Each processor will take care of moving data between cache and drives. The binfile you mentioned binds different slices together and allows backend processors to understand how your data is layed out on different drives, on different slices, ultimately allowing your frontend processors to read/write data on "logical devices" (that's how we call symdevs) :-)

Hope it's clear as mud .. ;-)

I'm pretty sure you'll have now more questions then before thus please post more questions :-)

12 Posts

October 20th, 2008 01:00

Thanks Stefano,

That cleared up quite a bit there.

So from a single disk, there can exist many hypers/slices with different RAID protection on it...this all controlled by the micocode i assume..

And the actual hardware on the SYMM which does this complex job is the processors, and with the help fm the binfile to remember where everyone is..

Is this a good summary from your expalantion? :)

Also, should a single HD fail(this wld mean a few hypers affected) and if there was a need to replace the drive, what actually tells the drive to perform a rebuild? And what RAID levels are the HD at??

Message was edited by: GAD
GAD

2 Intern

 • 

292 Posts

October 20th, 2008 06:00

Stefano, as usual, is absolutely correct. But I do want to make one further note, a hot spare will invoke when a drive fails only when there is a hot spare configured in the box and is available to be invoked. If there is no hot spare, you lose redundancy but code is smart enough to rebuild the data to recreate what was on the failed disk.

6 Operator

 • 

2.8K Posts

October 20th, 2008 20:00

I agree .. the DMX is amazing .. And Mike too ;-)

12 Posts

October 20th, 2008 20:00

Amazing stuff guys...

Thanks Stefano and Mike...question answered..:))
No Events found!

Top