IBM RS/6000 SP Handbook page 129

Description
The 7026-6H1 is packaged as a rack-mounted Central Electronics Complex
(CEC) drawer, cable attached to rack-mounted Remote I/O (RIO) drawers. The
CEC and I/O drawers offer redundant power options and redundant cooling. The
CEC drawer incorporates the system processors, memory, and supporting
systems logic.
System storage is provided via remotely attached SCSI, SSA, or Fibre Channel
storage subsystems. A fully configured 7026-6H1 server includes:
One 6-way processor card with 8 MB L2 cache per processor
32 GB of system memory
Two I/O drawers incorporating:
– 28 available hot-plug PCI slots
– Three available media bays
The 7026-6H1 CEC and I/O drawers are mounted in the IBM 7014 rack.
Additional space in the 7014 rack may be utilized for various storage and
communications subsystems.
AIX Version 4.3.3 or 5.1 software is included with every 7026-6H1 and can be
preinstalled.
The 7026-6H1 brings new levels of availability features and functions to the
Enterprise Server. The optional N+1 power subsystem provides redundancy in
case of power supply failure. Hot-plugged power supplies in the CEC and I/O
drawers allow concurrent repair when redundant power is installed. Optional
uninterruptible power supply (UPS) systems are supported for this system.
The 7026-6H1's CEC and I/O drawer cooling subsystems have N+1 fans, which
can be replaced concurrently if the optional redundant power is installed.
Systems equipped with optional redundant power have independent line cords
for each power supply in the CEC and I/O drawers.
The 7026-6H1 provides Chipkill Memory with selected memory options. Using
advanced technology derived from IBM's ultra-reliable System/390 enterprise
servers, Chipkill Memory protects the server from any single memory chip failure
and multi-bit errors from any portion of a single memory chip.
Memory chip failures can cause server system crashes that can result in the
permanent loss of business data. Many file server applications lack the means to
recover data after a catastrophic failure of this kind. Even when recovery is
possible, it can take hours, resulting in significant losses in productivity or
revenue.
Chapter 3. SP-attached servers and Clustered Enterprise Servers
105