topic Re: Does anyone know how 8.0cm is going to achieve SSI? in Network and Storage Protocols

Does anyone know how 8.0cm is going to achieve SSI?

daniel_fedick — Thu, 05 Jun 2025 07:21:08 GMT

I get that there is a global namespace for NFS and for CIFS and that all the systems are sharing a global IP across all the systems.

In the "Data ONTAP 8.0 Cluster-Mode" white paper there is a statement that says the following:

"By offering a single system image across multiple storage nodes, the global namespace eliminates the need for complex automounter maps and symbolic link scripts."

Are they saying that all controllers in the cluster have access to all the disks? If so, are we now creating a SAN between all the controllers and storage shelves in the cluster?

My assumptions are that each storage system has access to it's disks and the "Cluster-Mode" is only a statement in regards to the front end (I.e. NFS / CIFS Global Namespace.)

Does this sound correct?

Re: Does anyone know how 8.0cm is going to achieve SSI?

adamfox — Thu, 19 Nov 2009 19:35:48 GMT

First off, if you really want a deep dive on cluster mode, I strongly suggest you have you local team come in and give you one. They should be able to give far more detail than is feasible here.

That being said, each controller in the cluster has access to data on all of the other controllers through a back-end interconnected network. So if you have 8 controllers in your cluster, you can access data from all 8 from any one node in the cluster, thus giving you front-end scalability. This is particularly important when you consider high-performance aggregates (i.e. aggregates which are striped across multiple cluster controllers).

Like I said, there's plenty of deep dive info available on this, but I wanted to give you a quick, high-level answer.

Hope this helps.

Re: Does anyone know how 8.0cm is going to achieve SSI?

daniel_fedick — Fri, 20 Nov 2009 13:42:05 GMT

Will do.. I'll see if I can get into a deep dive up here. I need it I guess.

My assumptions are that you'd have to have a connection from every controller to every shelf in the cluster. In this case you'd need a fiber channel switch zoned for connectivity on the back end of some sort to allow ownership across all the controllers to all the disks.

I guess the real question is this: In a 7.x world, every disk has a primary and a secondary owner.

In the 8.x world, are there primary, secondary.. and tertiary etc.. owners on 1 disk? This to me would make it Single System Image.

What happens if you lose both controllers in 1 node of the cluster. Does node 2 have the ability to take control of those disks that were assigned to the first node?

I'm sure I'm botching this all up and will reach out to our guys up here to get some further wisdom in regards to 8.0 Cluster-Mode.

Thanks!