HP Integrity rx2600 Especificações Página 22

  • Descarregar
  • Adicionar aos meus manuais
  • Imprimir
  • Página
    / 26
  • Índice
  • MARCADORES
  • Avaliado. / 5. Com base em avaliações de clientes
Vista de página 21
These servers have high-availability power supplies, too. The rx5670, rx4640, and rx2600 power subsystems
provide high availability with N+1 redundant power options. The rx5670 comes standard with two hot-swap
power supplies, and an optional third supply can be ordered for 2+1 redundancy. The rx4640 and rx2600 each
come standard with a single hot-swap power supply; and an optional second supply gives these servers 1+1
redundancy of power supplies. To further enhance availability, each power supply has its own dedicated power
feed or line cord. Cords can be plugged into separate power grids for the maximum level of power protection.
Hot-plug disk drives
The rx5670 supports up to four SCSI disks, the rx4640 supports up to two, and the rx2600 supports up to three.
All disks are accessible from the front of the system and can be removed (or hot-plugged) while the server
continues to run.
Two dual-channel SCSI controllers manage the four internal hot-plug disks in the rx5670. For added availability,
disk pairs are on separate channels as well as separate SCSI controllers. This means that with disk mirroring, a
SCSI controller, SCSI channel, or root disk could fail and the server would continue to run properly.
A dual-channel SCSI controller manages the pair of disks in the rx4640. The disks can be configured either on a
single SCSI channel or one disk on each of the two channels with disk mirroring for added availability. When
only one SCSI channel is used for the disks, the second can be connected to an external device such as a tape
drive.
A single dual-channel SCSI controller manages the three disks in the rx2600. One channel links to two internal
disks; the second channel is connected to the third internal disk. This allows disk mirroring across separate SCSI
channels, further enhancing availability.
Multiple I/O channels
The multiple zx1 I/O channels in the rx5670, rx4640, and rx2600 provide failover, load balancing, and failure
isolation. In these servers, failures on one channel do not disrupt activities on other channels. Furthermore, both
servers deploy fully independent PCI-X buses to isolate traffic on I/O adapters. If a problem occurs on one
adapter, it will not interfere with traffic on another bus.
ECC and chip spare memory
The memory systems for the HP Integrity rx5670, rx4640, and rx2600 servers utilize error-correcting code to
correct single-bit errors as well as HP’s chip spare technology to protect against multi-bit errors.
Chip spare enables an entire SDRAM chip on a DIMM to be bypassed in the event that a multi-bit error is
detected on that SDRAM. In order to use the chip spare functionality, identical-sized DIMMs must be loaded
in quads. Different DIMM sizes are supported, as long as they are in different quads. For example, a quad of
512 MB DIMMs can be loaded along with a second quad of 1 GB DIMMs, and chip spare will be enabled on
all the DIMMs.
Because of the chip spare feature, the rx5670, the rx4640, and the rx2600 servers are completely resilient to all
SDRAM failures, regardless of the number of bits involved in the fault condition. This virtually eliminates memory
failures as a source of system errors.
Some other vendors deal with multi-bit SDRAM failures by accepting the fact that they will occur. That is, they use
a scheme that supports only failure detection, not failure correction. HP believes that this is unacceptable and a
dangerous choice for servers in business-critical environments. In fact, server systems that employ failure detection
but not correction are at high risk to fail due to memory problems.
CPU error correction and dynamic processor resiliency
In the HP Integrity rx2600, rx4640, and rx5670 servers, L1 and L2 caches both have full single-bit error checking
and correcting as well as double-bit error detection. Additionally, all the instruction and data paths also have
single-bit error-checking and -correcting capabilities. What’s more, the system processor bus has parity detection,
and the data path is covered by error correction.
The rx2600, rx4640, and rx5670 employ dynamic processor resiliency (DPR), too. With DPR, any CPU
generating correctable cache errors at a rate deemed unacceptable is de-allocated from use by the system. This
feature helps protect against a CPU degrading to the point where it may cause system crashes.
22
Vista de página 21
1 2 ... 17 18 19 20 21 22 23 24 25 26

Comentários a estes Manuais

Sem comentários