[cobalt-users] hdparm on a RAQ3

["Steve Bassi" <Steve@xxxxxxxxx>]

It was my belief that the RaQ is a machine designed from the bottom up
to optimally run Linux, and is pre-configured by Cobalt's experts, and
that consequently they would have done everything there is to do to maximise
disk i/o efficiency.

However, on my RAQ3

hdparm /dev/hda

/dev/hda:
  multcount    = 32 (on)
  I/O support  =  0 (default 16-bit)
  nmaskirq    =  0 (off)
  using_dma    =  1 (on)
  epsettings =  0 (off)
  nowerr       =  0 (off)
  readonly     =  0 (off)
  readahead    =  8 (on)
  geometry     = 19846/16/63, sectors = 20005650, start = 0

that would not appear to be the case. While multcount is active,  I/O
support and unmaskirq are not, and I would have thought that the RaQ could
benefit from them.

The question, then, is whether these options are not set because Cobalt did
not think about them, or are they not set because they know from experience
that there is no merit in them?

My immediate problem is safely testing any changes.

The man pages sensibly caution me to enter single user mode, but if I do
that then the system is normally restricted to the console and any network
sessions will be aborted.

Q1. Has Cobalt have configured the RaQ to enter single user on a remote link
[which is possible]?

Q2. Has anyone changed these seetings on a RAQ3, with any significicant
benefit.

Q£. Any comments or suggestions

Many thanks

Best Regards

Steve Bassi
London UK