Configuration Rules for a Mixed HP 9000 / Integrity Serviceguard Cluster, March 2007

Table Of Contents
HP Superdome Hybrid Servers
HP Superdome servers based on the HP sx1000, or the HP sx2000 chipset can support
configurations with both PA-RISC nPartitions and Intel® Itanium® 2 nPartitions in the same server. This
allows customers to transition from HP9000 to Integrity within a Superdome at the granularity of a
single partition. This provides customers the flexibility to transition each application to Intel® Itanium®
on their own schedule, reducing both risk and operational disruption.
There are a couple of general rules that apply to hybrid Superdome servers:
all CPUs in a cell must be of the same type; either all PA-RISC, or all Intel® Itanium® 2
all cells in a nPar must be have the same type of CPUs installed
a Superdome can consist of multiple nPars with each having a different type of CPUs installed
The following table gives an overview of the supported CPU / chipset combination for Superdome
hybrid servers.
Table 5. HP Superdome Hybrid Server supported CPUs
CPU sx1000 sx2000
PA8800 yes no
PA8900 yes yes
Intel® Itaninum® 2 single-core
processor with 9 MB cache (A1
stepping and A2 stepping) aka
Mad9m
yes yes
Montecito Dual-Core Intel® Itanium®
2 processor
no yes
Montvale no yes
Superdome hybrid servers don’t support the following CPUs:
PA8700 or older
Intel® Itanium® 1
Intel® Itanium® 2 single-core with 6 MB cache or less
Intel® Itanium® 2 multi-core; aka mx2
While certain hybrid Superdome configurations are supported with HP-UX 11iv1, Linux, and
Windows, only nPars with a minimum HP-UX release of 11i v2 UD2 are supported in a mixed
Serviceguard cluster. Please refer to the following white paper for more details on hybrid Superdome
servers:
http://docs.hp.com/en/6906/SuperdomeMixingRev1.3.pdf
Typical configuration examples
The most typical use case is an existing HP 9000 cluster to which a customer wants to add new
Integrity nodes. For applications that support mixed architecture failover, this process is straight
forward as long as the existing cluster hasn’t reached its maximum cluster size which is typically 16
nodes. Here is a list of some applications that are currently known to support mixed architecture
failover:
Oracle has granted support for 9i and 10g single instance databases.
IBM added support for Informix Dynamic Server 10.0.
15