"Kjetil Torgrim Homme" <ifi.uio.no> wrote in message
> simultaneous users? probably not.
> nntp.uio.no is a Netra T1. it has 50000 potential users, but usually
> number of simultaneous users is around 50, perhaps peaking into 200.
> it handles it fine, but it uses an A1000 for storage, pluss an
> external SCSI disk for overview
A1000 way to expensive and is 512MB RAM enough?
> no problem.
> : [rootreadme /]; df -k
> Filesystem kbytes used avail capacity Mounted on
> /dev/md/dsk/d0 1489367 1136993 292800 80% /
> /dev/md/dsk/d3 1489367 281659 1148134 20% /var
> /dev/md/dsk/d5 9272769 4798168 4381874 53% /news
> 1048576 88954 899700 9% /news/log
> 2097152 214854 1764753 11%[/ref]
Thanks for the df output, helps alot.
> how important is uptime and old news? in our setup, everything is
> stored redundantly except logs and overview. if (when) the overview
> disk crashes, we will experience ~12 hour downtime while it rebuilds.
> yes, but I like to have overview on a separate spindle.[/ref]
I realized i have a small UltraSCSI disk case that can hold 12 disks, that
should be sufficient.
On random I/O reads the Stripe and Raid-5 is comparable. On random I/O
writes the stripe is superior to Raid-5. What type of I/O operation is the
overview mostly used for? Is Raid-5 bad idea on the overview?