For Power5, the world changes.

As far as the DLPAR question, provided that HMC can resolve the LPARs
name (The name you see in WebSM for the LPAR) AND that the LPAR itself
has the same name for it's hostname (and is resolvable) you should have
the IBM.DRM daemon running in your LPAR. If you don't have the IBM.DRM
daemon running, starting it will be to no avail (startsrc -s IBM.DRM
will say that it's started but a subsequent lssrc will show it's down.)
To fix this read the end of this post. So you've found that IBM.DRM
is running on your LPAR, right-click the LPAR in WebSM....Hover over
"Dynamic Logical Partitioning"...Hover over "Processor
Resources"....click Add.

Now add the number of processors (or tenths) you wish to add.
Additionally, increase the number of Virtual Processors you wish to add
(there must be a minimum of one tenth of a processor available per
Virtual Processor.) More on this is a minute.

Click "OK" and your processor count will be increased provided there
are free resources in the pool.

Validate this by performing a vmstat (this will show you the lcpu -
"logical cpu * 2" or Virtual Processors assigned AND the ent "entitled
capacity")
# vmstat

System configuration: lcpu=4 mem=2048MB ent=0.20

kthr memory page faults cpu
----- ----------- ------------------------ ------------
-----------------------
r b avm fre re pi po fr sr cy in sy cs us sy id wa
pc ec
1 1 296823 135018 0 0 0 1 4 0 11 1796 358 3 2 96 0
0.01 5.8

The LCPU should equal the number of Virtual Processors times two
(unless you've turned off SMT). Entitled Capacity will show you the
amount of Processor capacity you've been assigned.


Capped/Uncapped

Cappped processors is a "safe" way to run your system. Uncapped though
is a "brilliant" way to run a system. Short story... Uncapping all of
the partitions on your system allows you to assign "priorities" to each
LPAR. The hypervisor then every 10 milliseconds analyzes the workloads
and priorities. This analysis looks at the "free pool" - the amount of
processor capacity currently idle across all LPARs and allocates the
amount free to each LPAR which "needs it." The ceiling of how much can
be allocated to a partition in this allocation is controlled by the
number of Virtual Processors assigned.

Example.... you have an Oracle instance which has been allocated 0.3
processor through one Virtual Processor. The Oracle server receives a
large query and jumps to 100% utilization (on 0.3 processors). In the
next 10 milliseconds, the hypervisor sees this partition struggling and
checks the LPARs priority (set at 200) against the priorities of the
other LPARs "in need." The hypervisor might find 1.3 processors of
idle capacity in other LPARs and allocated the Oracle server 0.7
processors. (It can't exceed one full processor as you've only
allocated one Virtual Processor.) If there had been two virtual
processors, the Oracle server may have been granted the entire capacity
of the free pool (1.3 processors) across it's two processors. Let's
say that in the next 10 milliseconds, the systems which from which that
capacity was "stolen from" need their capacity back - it'll be granted
back automatically.

SO.... be careful. Uncapping everything in your system can work well
provided you are careful in your planning.

Good Luck,

Chris Young
Consultant for Hire
614-804-UNIX