md3000i and snapshots?

[Harald_Jensas at Dell.com]

> -----Original Message-----
> From: linux-poweredge-bounces at dell.com [mailto:linux-poweredge-
> bounces at dell.com] On Behalf Of Scott R. Ehrlich
> Sent: 29 July 2008 13:10
> To: Parrott, Nick
> Cc: linux-poweredge-Lists
> Subject: RE: md3000i and snapshots?
> 
> Hi Nick:
> 
> I appreciate that feedback, but to get an additional head-start on this
> before reading the referenced URLs, how do I really _know_ how much to
> allocate?  

How much space you need to reserve for SnapShots depends on your environment. 
 - How often do you plan to take SnapShots? 
 - How many active snapshots at the same time?
 - For how long will SnapShots be active? 
 - How much does the data in your source VD change in the time your snapshot is active?

Example:

- You create a 500 GB Virtual Disk.
- Your will create a SnapShot of the 500 GB VD every 12 hours. 
- You will only have 1 SnapShot at any given time. (E.g after 12 hours has passed, you delete the SnapShot and create a new SnapShot at that point in time.)
- You know that the data change rate for your 500 GB VD is MAX 7% in a 12 hour window. (500 GB * 0.07 = 35 GB)

In this example you will need at least 35 GB of free space that SnapShot can use as SnapShot Repository.

Need I say that in most cases the change rate will grow?  If you calculate 7% change rate on the source VD go for something like 15% reserved space.

> I want to make as much capacity available as possible for
> users.  If I have 6 x 300 GB drives + 1 hot spare, how many disks should I
> create in a virtual group if I want a RAID 5 set?   Would I use all 6, but
> then SS will claim no space is available for it to use?   Do I use 5 disks
> for the RAID 5, and the 6th for SS?   Maybe someone wants their own
> project set, thus I'd have multiple RAID 5 volumes, or a RAID 5 volume and
> a RAID 1.

Dell recommends to configure a Disk Group, then create 1 or more Virtual Disk that does not use the entire capacity of the Disk Group. E.g leave capacity at the "end" of the Disk Group for SnapShot repository for the VDs in the Disk Group. 

RAID Levels, number of drives and how you divide your 6 available drives depend on your environment. Performance requirements, space requirements etc.

That said, in your case with 7 drives, 1x Disk Group with 6 Drives configured as RAID 5 + 1x HotSpare is most likely a good idea. You can then divide the 6 drive RAID 5 Disk Group into several Virtual Disks and assign one VD to users and another VD to a specific project if you need too.


> 
> I need some good, specific documentation that details how your SS
> implementation works, so I can truly appreciate how to properly set it up
> on my end.  Maybe the URLs give that info.  I'll find out later today.
http://support.euro.dell.com/support/edocs/systems/md3000i/en/UG/HTML/snapshot.htm#wp1001413

> 
> How does Dell's snapshot technology actually work?   I know it takes a
> live moment-in-time image, but what is that image actually based on? Is
> the data compressed?   Does it perform a incremental or date difference or
> something else method?   Are they full duplicate data, or some method of
> pointers to data living somewhere else, since data does change.
> 

The snapshot technology use COFW (Copy On First Write). When a snapshot of a source virtual disk is created any writes to the source VD is copied to the snapshot repository before the write to source VD is performed. The technology does use pointers, when the snapshot is first created all pointers point to the source VD. But as writes are performed on the source VD, and data is copied to the snapshot repository the pointers are also changed so that they point to the 'original data' that has been copied to the snapshot repository.



--
Harald Jensås