Controller ownership: backend disks and data - new architecture and/or implementation

Hello Product and Development teams,

Over Latin America we faced some situation where customers did see (a lot) value in solutions where multiple controllers can work together, to minimize downtime, to achieve some workload requirements and so on. This requirement is normally achieved today by vendors that offer active-active symmetric architecture.

Our current architecture, thinking on SAN Volume Controller for example, each I/O Group have the ownership of their backend capacity, therefore, if for some reason both controllers do fail, all data managed by this I/O group become unavailable (until problem is fixed). The PBHA or even Hyperswap imply to have some limitations to the environment or even cluster/scale-out.

My main idea here, after some discussions with colleagues and customers, is to propose an idea to IBM have a solution that can desegregated the backend to be managed by only one I/O Group.

One example, a future release of SVC have the ability to share with some other SVC the same backend in a way that if an I/O group entirely fail, the other I/O Group still have the ability to access data transparently and without to implement any other topology, like PBHA. If customer need more capacity, more shelves of disks can be installed, if customer need more processing power, more controller (I/O groups) can be installed.

According last report of Magic Quadrant of Gartner, HPE, one of leaders of the report, have this implementation or function as a strength - Gartner: "HPE’s four-node 32-core CPU Alletra MP SDS disaggregated architecture allows customers to cost-effectively scale platform capacity and performance independent of each other, with a 100% data availability guarantee."

Thinking that IBM will have new products some point in the future (GEN5 platform and EDSFF disks) and I believe the next gen platform will have NVMe disk expansion, maybe we can use this moment to bring this feature to IBM solutions and combine with all other impressive features and functions that we do have.

Thanks.