Cell 4 is connected to IO chassis 0-0-1. The Slot- ID for the Fibre Channel card would therefore be 0-0-1-6.
The associated HP-UX hardware path would be 4/0/14/0/0.
Slot 6 is an appropriate slot for this card. Slot 6 is a quad-speed card offering approximately 530 MB /second throughput. A 2GB Fibre Channel card requires a throughput of 2GB/8 = 256MB/second. A quad-speed slot is more than capable of providing this level of IO performance.
All other information in the question is spurious and designed to divert the reader from the actual question.
A2:
The three Single Points Of Failure in a Superdome complex are:
The System backplane
The UGUY board
The HUCB board
Technically, there is nothing we can do to eliminate an SPOF completely without providing a truly fault-tolerant solution. Superdome is not fault-tolerant. In order to alleviate as much downtime as possible should an SPOF actually cause the complex to fail, we could employ a second complex and utilize software such as HP's ServiceGuard where we could configure individual partitions to be members of a high-availability cluster. If a complex fails (due to the failure of an SPOF component), a partition in another complex, belonging to the same cluster could undertake the running of affected applications.
A3:
The parmodify command was used without the “B option. This option will instruct the GSP to boot the partition once all cells (according to the new SCCD ) are at BIB . The new SCCD can only be pushed out to cells that have BIB set and are inactive. Because the “B option was not used, the new SCCD will be pushed out to the affected cells but will remain spinning on BIB . The administrator will have to issue the GSP command BO in order to manually boot the partition past BIB .
A4:
Use the nifty-54 diagram to construct the following supported partition configuration (in order):
Finance: cells (cabinet 0) = 0, 1, 2, 3, 5, and 7 (0 and 2 are both connected to an IO chassis).
Sales: cells (cabinet 1) = 0, 1, 2, and 3 (0 and 2 are both connected to an IO chassis within this cabinet).
IT: cells (cabinet 1) = 4 and 6 (both are connected to an IO chassis in this cabinet).
Marketing: cells (cabinet 1) = 5 and 7 (both are connected to an IO chassis in an IO expansion cabinet).
Research: cell (cabinet 0) = 5 (connected to an IO chassis in an IO expansion cabinet.
Notes:
The Finance partition is the largest and is assumed to be of major importance to the business. It has been housed in cabinet 0 as per the nifty-54 diagram. The only other partition in cabinet 0 is the Research partition, which is seldom used and will cause little impact to the performance of the Finance partition. Housing a more active partition, e.g., IT or Marketing, in cabinet 0 may impact the performance of both partitions when both partitions need to access other cells across the XBC interface (although the XBC has adequate bandwidth to accommodate IO from every Cell Controller attached to it). Another reason for housing the Research partition in cabinet 0 is that it leaves a cell free in case the Finance partition needs to be expanded. In such a situation, it is best if the entire partition is housed in the same cabinet.
The Research partition has no High Availability feature in case an entire cell fails. This has been noted and accepted by the Research department.
It is assumed that an IO expansion cabinet is available because the question says that all cells are connected to an IO chassis. This is currently not possible without the use of an IO expansion cabinet.
A5:
The cell-based architecture implemented by Superdome introduces different memory access times when a partition is accessing memory from different cells on different XBC interfaces and in different cabinets . This is a classic feature of the Non-Uniform Memory Access ( NUMA ) architecture. In its initial release, HP-UX 11i version 1 does not make any use of this feature and simply interleaves memory access across all cells in the partition evenly. This alleviates any single latency by utilizing the memory bandwidth across all cells in the partition. HP-UX 11i version 1 "views" an nPar as simply an SMP server. HP-UX 11i version 1 will maintain cache coherency across all processors in the partition. HP-UX 11i version 2 starts to utilize the NUMA aspects of Superdome by allowing the administrator to configure Cell Local Memory whereby a proportion of memory interleaving is not performed. This has been seen to even further improve application performance in specific situations. Cache coherency is still maintained across all processors in the partition, hence encapsulating all necessary features of the cc- NUMA architecture.
