Acer SmartRAID V Bedienungsanleitung Seite 170

  • Herunterladen
  • Zu meinen Handbüchern hinzufügen
  • Drucken
  • Seite
    / 256
  • Inhaltsverzeichnis
  • FEHLERBEHEBUNG
  • LESEZEICHEN
  • Bewertet. / 5. Basierend auf Kundenbewertungen
Seitenansicht 169
7-2
Theory of Operation
Intelligent RAID Controllers
File servers achieve maximum throughput when the performance of their
component subsystems is optimized for the CPU bandwidth. If any
component is not capable of supplying data at an optimum rate, that
component will restrict the data path performance and slow down the
entire system. Conversely, a component with significantly greater
performance than the rest of the system can have much of its
performance potential unused. Any attempt to speed up a server by
increasing the performance of only some system components usually
results in a negligible overall increase.
The speed of CPUs used in currently available file servers has increased
greatly over those available a few years ago. However, the performance of
data storage subsystems has not kept up. As a result, many servers have
data storage subsystems that are not capable of supplying data to the CPU
fast enough. Consequently the CPU can be idle for significant periods
waiting for data and these servers cannot perform at their full potential.
In a traditional server, the host CPU handles the processing of I/O
interrupts from peripherals, disk storage and network devices. This design
was developed for the single-user and non-multitasking operating systems
which were in use when microcomputers first became widely available.
However, this design becomes a performance restriction when
incorporated into graphics workstations or multi-user servers that have
pipelined-architecture processors. On a busy server, the large amount of I/
O interrupts sent to the host CPU reduces the number CPU clock cycles
available to process application software instructions. Additionally, each
interrupt causes the CPU internal pipelines and caches to flush their data,
which results in additional CPU performance degradation.
Seitenansicht 169
1 2 ... 165 166 167 168 169 170 171 172 173 174 175 ... 255 256

Kommentare zu diesen Handbüchern

Keine Kommentare