Highlighted
jnickel
1 Copper

Celerra concepts and information

Jump to solution
Is there any documentation that explains a few things? Mostly concepts.
I have read all the documentation I could find and I understand that a Data Mover is a Blade - it literally moves the data between the disks and the outside servers.  The Control station is also a Blade, but it runs the Unisphere and handles the management of the Data Movers, Disks and more.  Please correct me if I am wrong.
However....some things are not very well explained:

1) What is the difference between a RAID Group, Storage Pool and Volume and the relationship between them? It seems that if I want to use all the space, I am best to use the "default" or pre-made Storage pools, and it seems like it makes the Volumes automatically....I am very confused.
2) How the automatic or pre-defined storage pools assign space? can you remove the space from the automatic ones and assign it to your own pools?
3) How space is added - when I added a 300 GB VMDK to the Celerra Uber VSA as an example, split it up into a bunch of "disks"...is the idea that I added a "DAE" or shelf?
4) In a real Celerra, do you need to "provision" disks before they can be used?

I thought I understood this stuff - but I am confused.

I thought a RAID Group was simply a collection of disks into a RAID level (1, 5, 6, 10 etc). I was expecting to be able to create RAID groups of whatever size I chose - a 7 disk RAID 5, or a 6 disk RAID 1 etc.

I thought a Volume was a collection of RAID Groups.

And I thought a Storage Pool was a collection of Volumes....

Please clarify for me if you have the time.

Thanks!

Jim
1 Solution

Accepted Solutions

Re: Celerra concepts and information

Jump to solution

You’ve got the hardware correct. Each datamover is also a blade, also called an xBlade.

Starting from the disk and going up…

Physical Disk

RAID Group (collection of physical disks)

LUN (block volume carved from a RAID Group, typically 2 or more from each RAID Group when using Celerra, evenly split among the two SPs)

Celerra Volume (celerra representation of LUNs)

Storage Pool (a collection of volumes)

Filesystem (volume carved from a storage pool)

There are actually several types of volumes on a Celerra and some types are made from others…

Disk Volume = equivalent to a LUN presented from the backend storage to the datamover

Slice = used to break a volume into smaller pieces which can be used for multiple filesystems

Stripe = a collection of volumes on which data is striped evenly for good performance

Meta = a collection of volumes combined together by concatenation

When using AVM (Automatic Volume Management) I believe the default storage pool on a Celerra is just a Meta of the available disk volumes. Filesystems are allocated from the individual LUNs(Disk Volumes) in the pool as they are created in a sort of round robin fashion. Performance is good but read/writes do not necessarily hit all disks in the pool.

If you use MVM (Manual Volume Management) you can build a storage pool using a combination of the above listed volumes. Depending on your requirements, you can get better performance using MVM. For example, if you had 2 LUNs per RAID Group, you could create two Stripe Volumes, each made from 1 Disk Volume per RAID Group. Then concatenate the two Stripes into a single Meta Volume. You can create filesystems from the Meta directly or you can create multiple Slice volumes from the Meta in order to provide more granular allocation of the space to the filesystems.

Typically, AVM provides adequate performance for 95% of workloads.

I hope that helps.

3 Replies

Re: Celerra concepts and information

Jump to solution

You’ve got the hardware correct. Each datamover is also a blade, also called an xBlade.

Starting from the disk and going up…

Physical Disk

RAID Group (collection of physical disks)

LUN (block volume carved from a RAID Group, typically 2 or more from each RAID Group when using Celerra, evenly split among the two SPs)

Celerra Volume (celerra representation of LUNs)

Storage Pool (a collection of volumes)

Filesystem (volume carved from a storage pool)

There are actually several types of volumes on a Celerra and some types are made from others…

Disk Volume = equivalent to a LUN presented from the backend storage to the datamover

Slice = used to break a volume into smaller pieces which can be used for multiple filesystems

Stripe = a collection of volumes on which data is striped evenly for good performance

Meta = a collection of volumes combined together by concatenation

When using AVM (Automatic Volume Management) I believe the default storage pool on a Celerra is just a Meta of the available disk volumes. Filesystems are allocated from the individual LUNs(Disk Volumes) in the pool as they are created in a sort of round robin fashion. Performance is good but read/writes do not necessarily hit all disks in the pool.

If you use MVM (Manual Volume Management) you can build a storage pool using a combination of the above listed volumes. Depending on your requirements, you can get better performance using MVM. For example, if you had 2 LUNs per RAID Group, you could create two Stripe Volumes, each made from 1 Disk Volume per RAID Group. Then concatenate the two Stripes into a single Meta Volume. You can create filesystems from the Meta directly or you can create multiple Slice volumes from the Meta in order to provide more granular allocation of the space to the filesystems.

Typically, AVM provides adequate performance for 95% of workloads.

I hope that helps.

jnickel
1 Copper

Re: Celerra concepts and information

Jump to solution

That is extremely helpful!

Thank you.

When you say "LUN", I always think of the LUNs that I present to my ESX servers, but in this case, it would appear you are referring to LUNs presented to the Control Station blade?  But these are hidden - you can only see the RAID Groups and the Volume - but I think you are saying that the Volumes are the same as the LUNs at this level?

When presenting LUNs (iSCSI LUNs for example) to my ESX servers - does this happen from the Filesystem level, the Volume Level or the Storage Pool level? Or can it happen from mulitple places? From your explaination, it sounds like it happens at the Pool level.

If I wanted to only use MVM, can I take disks away from the default storage pools and make my own?  When I attempted this before, I couldn't seem to get full disk usage....but maybe that is because of the Vault area?  I understand that the first 4+1 set of disks is special and used for the "Vault" - for the storage to store data about itself....is that correct?

One last thing - if there are 2 Datamover xBlades, then each should be assigned it's own IP and will have access to the same storage?  This will result in an additional 2 paths to the same LUNs presented to my ESX servers?

Thanks again for the quick response!

Jim

0 Kudos

Re: Celerra concepts and information

Jump to solution

The LUNs I was referring to are not visible to you in the Celerra management tools, but would be represented as Disk Volumes. If you logged directly into the backend array (ability to do this is dependent on the system model and licenses) you would see raw RAID Groups and LUNs. And the datamovers/blades are represented just like a host would be if attached to a Clariion.

iSCSI LUNs are built inside a Filesystem. So you create a filesystem in a pool, then create an iSCSI LUNs

(ESX Datastore->iSCSI LUN->FileSystem->Pool/Volume-> Disk Volumes->Backend LUNs->RAID Groups->Physical Disks)

The control station does not see any of the storage directly. The datamover blade is directly connected to the backend storage and sees the storage space that has been assigned to the Celerra. The Control Station is a 1U server running Linux which is used for management purposes only. A failure of the Control Station does not affect access to Celerra filesystems or iSCSI LUNs. Aside from running the webserver for Celerra Manager or Unisphere, the ControlStation monitors the blades and handles failover in the event of a blade failure.

Data Flow is like this…

iSCSI: Host -> Datamover/Blade -> Backend Storage

NAS: User -> Datamover/Blade -> Backend Storage

Management: Admin User -> Control Station -> Datamover/Blade

The first 5 drives on the system contain the FLARE code which is the backend block storage operating system AND the DART code which is the datamover operating system. The presence of that software consumes quite a bit of space on those drives, reducing the available space to user data.

Ability to take volumes away from an existing pool and reassign to a new pool or volume depends on whether there are existing filesystems in the pool. Explaining all the nuances of how to do this would be very difficult in this type of forum and would best be handled through EMC Support.