We have 27 instances to be clustered on 2-node Extended RAC (Inter-DC distance is 60 km on DWDM) clusters. There are 6-8 RAC instances per BL460 G8 Blade used as RAC nodes. Storage is ASM (Grid 188.8.131.52) and on HP P9500 SAN. Each Data-Center will have 5 c7000 enclosures hosted each having 12-16 BL460 Gen 8 blades(also 4+4 BL680 G7 blades). Each Blade will have 2 SAN ports in Mezzanine slot-1, each having 8Gb/s (Giga-Bit) total i.e. 4Gb (Primary) and 4Gb(Secondary).
Per standard Extended RAC configuration, there will be Production Network, RAC Interconnect (Cache Fusion) networks and the SAN networks routed through VLANs through the Chassis Backplane into the SAN FC Switches. Flex-10 fabric will be used. The database sizes hosted range from 200 GB to 18TB. Blades having 7-8 RAC instances each will host smaller databases from 300-700 GB. All disk allocation is in HP SAN P9500 Storage and the blades are bare-metal (with O/S and Oracle 11gR2 RAC software only) and without local storage (Disks). Boot from SAN will be used.
So, will 8Gb/s bandwidth capacity on each HBA/SAN port (in Mezzanine Slot 1) – be enough to route and handle I/O of the multiple 6-7 SAP Instances and OLTP/DSS application databases?? I understand of course this depends on the type of the applications, their acceptable IO throughputs/SLAs etc. To my knowledge, in most of standard enterprise Blade-System c7000 enclosure deployments 8Gb/s is standard? I understand this needs definitely to be tested for sure(and we are planning to do acceptance/stress testing too). But I am looking for empirical numbers to do sizing.
Please can you let me know your thoughts/inputs.. for any empirical sizing numbers..
Need to confirm back to Blade deployment team soon..