IBM 7014 Model T00 rack with eight pSeries
630 Model 6C4 drawers
To further maximise system
availability, the pSeries 630 server
has built-in fault and error correction
functions. For main memory, Chipkill
memory technology – developed by
IBM for mainframe servers – detects
multiple bit errors and corrects most
of them transparently. If the error
rate exceeds the critical threshold,
a maintenance action is initiated
automatically by the system, to be
resolved at the user’s convenience.
IBM studies indicate that IBM
systems with Chipkill memory are up
to 100 times less likely to experience
an outage due to memory failure.
5
In addition, redundant, spare main
memory chips are provided. Through
a technique known as bit steering,
these spares can be dynamically
activated and replace a failing
memory chip in the event multiple bit
memory errors exceed a threshold.
The use of advanced memory
technologies – ECC, Chipkill and bit
steering – helps protect the server
from memory failures that can cause
costly, unscheduled downtime. These
automatically invoked functions are
part of the innovative IBM
^
architecture.
pSeries 630 servers also feature the
ability to deallocate critical system
resources, including the processors
and L2/L3 cache (on four-way
systems) and PCI-X bus slots. In
the unlikely event that one of these
components fails or indicates an
impending failure, this capability –
working with the AIX operating
system and service processor – is
designed to dynamically take the
faulty component offline. Its workload
is reassigned automatically to other
resources such as processors and
memory, to avoid interruption. If the
system must be rebooted, previously
deallocated components will
not be included, therefore avoiding
repetition of the error condition.
Replacement of the failing component
can be scheduled during normal
service to minimise system and
application downtime.
Reliability and availability features
also include optional redundant
hot-plug power supplies and cooling
fans, which can be easily replaced
without affecting system operations.
Also available are environmental
monitoring functions such as
temperature monitoring that increases
the fan speed in response to
above normal temperatures.
For near continuous availability, from
two to 32 pSeries 630 servers can
be clustered with High Availability
Cluster Multiprocessing (HACMP)
software from IBM. HACMP helps
to minimise downtime of systems
and applications, providing a superior
base for high availability. This is
an essential ingredient of business-
critical environments.
Great packaging
The pSeries 630 offers the same
electronics in two different packages
for configuration flexibility.
The Model 6C4 is an industry-
standard 470 mm four Electronic
Industries Association Unit (4U)
drawer, designed to provide
maximum power in a ‘rack and stack’
environment. It can easily be installed
in an existing IBM or 470 mm original
equipment manufacturer rack. Up to
nine Model 6C4 servers may be
installed in an IBM 7014 Model T00
36U rack.
The Model 6E4 is a compact
deskside unit, just 172 mm wide
and 609 mm deep and fits ideally
in an office environment.
