iSCSI design best practice with multiple HP Storage Arrays

[pjapk]

I'm in the process of designing a config which will have VMware storage on one iSCSI array (P2000 G3) and B2D storage on another (MSA2000i).

As both are Dual-controller iSCSI devices, best practice suggests a two subnets per array (A0/B0 in one subnet, A1/B1 in the other) and non-routed.

Given a requirement for one server to access both arrays, this would potentially suggest at least 4 x NIC ports for iSCSI only in this server (with a further 2 ports minimum per additional iSCSI array).

Clearly this isn't scaleable so I'm trying to determine where the compromise should be made - either using just two subnets for both arrays (i.e. A0/B0 for BOTH arrays in one subnet & A1/B1 for BOTH arrays in the other, then a single NIC in each subnet at the host end).

Or, whether the SP's should remain in different subnets, but instead have them routed (again, resulting in a single host NIC port in each subnet).

I've been googling extensively & not found any guidance for this scenario so would appreciate any advice (or pointers to existing info).

My prior experience is with smaller-scale FC storage so this is the first time I'm delving deeper into iSCSI...

Regards,

Paul

[pjapk]

Just to clarify, the plan is to have Read-Only access to P2000 using Veeam in direct-san mode to back-up the VMware data to the MSA2000, so, looking to "read" with one NIC on the backup server and "write" with the other to maximise throughput.