Acer SmartRAID V Manuel d'utilisateur Page 192

  • Télécharger
  • Ajouter à mon manuel
  • Imprimer
  • Page
    / 256
  • Table des matières
  • DEPANNAGE
  • MARQUE LIVRES
  • Noté. / 5. Basé sur avis des utilisateurs
Vue de la page 191
7-24
Theory of Operation
Command Queuing
I
2
O messaging allows the controller to queue multiple commands from the
host and execute them out of order for greater efficiency. When a
command is complete, the controller transfers status information to host
memory. This status message contains information indicating which
command was just completed and enables commands to be executed in a
different order than they were issued by the host. This allows commands
to be queued and rearranged so that they are executed in a more efficient
order for the particular peripheral configuration.
SmartRAID V controllers also support Tagged Command Queuing on the
peripheral bus. This feature allows the controller to transfer multiple
commands to a single device if that device can accept queued commands.
Auto Request Sense
When a SCSI error occurs, the controller automatically issues a Request
Sense command and returns the data to the host in the message reply
frame. Data returned by this command is stored in the controllers Event
Log. You can use Storage Manager to view the Event Log.
Byte/Word Alignment
Data to be read or written by the SmartRAID V controllers need not be
aligned to word or long-word boundaries in system memory. If the data
starts on an odd memory address, the necessary bus control signals will
be generated by the SmartRAID V to access 8-bit, 16-bit, 24-bit and 32-bit
data as required. Data blocks of any byte length are correctly handled by
SmartRAID V.
Scatter/Gather
SmartRAID V controllers support Scatter/Gather. This is a method of
providing multiple memory addresses for data transfer in one host
command. This ability greatly increases performance in environments
such as UNIX, Novell NetWare and Windows NT. Scatter/Gather is
necessary because when using virtual memory addressing schemes,
system memory that may appear contiguous to the user can actually be
fragmented into many widely scattered physical address locations.
Because of this, when accessing a large amount of contiguous data from a
peripheral device, it is often necessary to split the transfer into many
different locations in system memory.
Vue de la page 191
1 2 ... 187 188 189 190 191 192 193 194 195 196 197 ... 255 256

Commentaires sur ces manuels

Pas de commentaire