Steve,

iSCSI is NOT a "replacement" for SnapManager for Exchange.

iSCSI will let you treat a file on a filer as a block device. There ends its usefulness. In other words, without host-side tools like SnapDrive or SnapManager you aren't any better off than with a dumb local block device (except that writes will be nice and fast due to block indirection).

There are three levels of data consistency:

1. Block level - this means that a physical block reported to have been written to disk has or will make it there intact. This is your typical DAS or SAN-level consistency.

2. Filesystem level - this means that all the blocks required for a consistent point-in-time image have been written successfully when the filesystem manger considers this to have been done. SnapDrive is NetApp's volume management plug-in for VLD, iSCSI and FCP "block" devices that manages this on a Windows host where that host is managing a NTFS volume laid down on the virtual block device. What happens is: you tell SnapDrive on the server to take a snapshot - it then tells the windows NTFS filesystem manager not to accept any more file I/O requests and to complete any outstanding I/O (flush writes) - then and only then does it tell the NetApp to take a WAFL-level snapshot of the volume hosting the virtual disk. It then tells the NTFS filesystem manager on the server that it can start processing requests again. If you simply took a NetApp level snapshot of the volume holding the virtual disk, then the image of the virtual disk could internally be in an inconsistent state, similar to a dirty shutdown of the Windows server.

3. Application level - this means that all of the file I/O OR block I/O operations required to keep an applications data files consistent from an application perspective (bearing in mind that the application will be managing outstanding operations in RAM) are completed at the times when the application expects, such as when you shut the application down or when you send a request through it's backup API to backup or restore it's data.

Raw iSCSI gives you level 1 consistency only. There is no facility for taking consistent point in time snapshots that you can use for online backups. You can use filer level snapshots if you like, but there's arisk that you won't be able to mount the snapshotted volume without some corruption.

SnapDrive gives you level 2 consistency. This is better and works will with databases that have a hot backup mode, like Oracle, or where you don't mind halting the database for a few seconds so that you don't have to deal with log files. To use this with Exchange you still have to manually manger restores with integrity checks etc.

SnapManager provides Exchange-specific functions through the Exchange backup API to allow you to use snapshots to manage online backups etc. This gives you level 3 consistency for your Exchange backups as well as super fast restores from snapshots etc.

iSCSI is good, but using it by itself is like taking the VLD component of SnapDrive or SnapManager without the rest.

-----Original Message----- From: Jeff Burton [mailto:burtonj@pprd.abbott.com] Sent: Friday, 28 February 2003 8:57 AM To: Steve Evans Cc: toasters@mathworks.com Subject: Re: iSCSI Buzz

On Mon, Feb 24, 2003 at 10:12:44AM -0800, Steve Evans wrote:

I've been listening to all the iSCSI buzz with NetApp in the news. From what I've gathered all you need is OnTAP 6.4 and a $700 adapter for the host server.

What are people planning on doing with this. I would like to use it with Exchange 2000 and not purchase the SnapManager for Exchange option (at $20,000).

Steve Evans SDSU Foundation (619) 594-0653

Per: http://now.netapp.com/NOW/products/iSCSI/

you also need a license for iSCSI, but it sounds like it's free as long as your filer's software support agreement is active. I was surprised to find the license free, so I may be missing something.

We do not have any plans to use iSCSI anywhere in our environment anytime soon. I seem to remember Gartner Group stating that iSCSI will not become popular until late 2004 or later.