Table of Contents
Advertisement
Error handling. Parity and other error conditions are detected
on the PCI buses. The memory checking scheme corrects
single-bit errors and detects double-bit errors. Multiple ECC
corrections to single-bit errors detected by the operating
systems help in determining where in the system the error
originated. Errors are logged for analysis.
Disk hot swap. The hardware is designed to enable hot swap
of disks. Hot swap is the removal of a disk while the rest of
the system remains powered on and continues to operate. This
feature contributes significantly to system availability. Since
many disk problems can be fixed without shutting down the
entire system, users lose access only to the disks that are
removed.
N+1 power redundancy. A second or third power supply can
be added to provide redundant power to the chassis. A second
power supply is needed for more than two CPUs or if a second
disk cage is installed. In this case the third supply provides
redundancy. The third power supply provides full N+1 redun-
dancy for configurations using up to 24 Gbytes memory.
Power supplies are 720 watts (DC). Each has two LEDs to
indicate the state of power to the system.
An external UPS can be purchased to support critical customer
configurations. Because power is maintained for the entire
system (CPU, memory, and I/O), power interruptions are
completely transparent to users.
Installation and Maintenance
The systems are designed for easy hardware, software, and
option installation. Options ordered with a system are pre-
installed and tested at the factory. The operating systems are
also installed at the factory.
Additional CPUs, memory, power supplies, and disks can be
added to the tower and pedestal systems by anyone with
appropriate technical training and experience. Installation of
components in a rackmount system is reserved for service
providers and self-maintenance customers.
Clustering
A cluster is a loosely coupled set of systems that behaves (is
addressed and managed) like a single system, but provides
high levels of availability through redundant CPUs, storage,
and data paths. Clusters are also highly scalable; that is, CPU,
I/O, storage, and application resources can be added incre-
mentally to efficiently increase capacity. For customers, this
translates to reliable access to system resources and data, and
investment protection of both hardware and software.
Clustering allows multiple computer systems to communicate
over a common interface, share disks, and spread the comput-
ing load across multiple CPUs. Clustering is implemented
using our traditional interconnects and using the newest tech-
nology.
PCI to Memory Channel Interconnect
Under Tru64 UNIX and OpenVMS, you can build high-avail-
ability clusters using the PCI to Memory Channel intercon-
nect. The Memory Channel interconnect is a high-bandwidth,
low-latency PCI-based communications interconnect for up to
eight AlphaServer systems. Data written to one computer's
memory is shared by other computers on the Memory Channel
bus.
The PCI adapter is the interface between a PCI and a Memory
Channel bus. This bus is a memory-to-memory computer
system interconnect that permits I/O space writes in one
computing node to be replicated into the memories of all other
nodes on the Memory Channel bus. A write performed by any
CPU to its reflected address region results in automatic
hardware updates to memory regions in other nodes. One
node's write is "reflected" to other nodes as a direct side effect
of the local write. This provides a memory region with
properties similar to a high-performance shared memory
across a group of nodes.
Operating System Support
For clustered Tru64 UNIX systems, TruCluster Software
solutions allow users access to network services and provide
further failover recovery from server, network, or I/O failures.
Tru64 UNIX cluster systems use the SCSI bus and/or PCI to
Memory Channel interconnect bus between disks and systems.
OpenVMS cluster systems use the CI, SCSI, Ethernet, FDDI,
and Memory Channel as the interconnect between disks and
systems.
The primary means of clustering AlphaServer ES40 systems
depends on the operating system.
CI clusters, OpenVMS only
Memory Channel, Tru64 UNIX and OpenVMS
SCSI clusters, Tru64 UNIX and OpenVMS
11
Advertisement
Table of Contents
