did you observce cpu utilization and/or load average too?
do you think SCSI in the master will help responsiveness?

in my case it looks like the cpu needs to:
*sycrhonize drbd (postgres data + indexes)
*take care of interrupts for 4 lan cards (heartbeat, drbd sync, backbone,
and ltsp local net(x over lan))
*take care of all (4) X desktops and X apps (because of ltsp)
*ssh vpn

hmmm...

Ariz Jacinto said:
> we've evaluated the same technique before but
> aborted the plan due to the storage's single
> point of failure. instead, we're now using
> DRBD's (www.drbd.org) Raid 1 over LAN replication.
>
> the replication speed is impressive enough. it
> is at 20Mbytes/sec (yup, Bytes & not bits) max
> on a typical 7200 rpm HDD & 100Mbps switch but
> using a lean linux kernel.
>
> using a disk mirroring technique for database
> replication should be your last resort since a
> lot time will be wasted during a full-sync
> instead of only updating the necessary tables,
> rows,etc. when using a typical db replicator
> not to mention the absence of query load-balancing.
> but there's no doubt that it is effective than
> those replicators that are not production-ready
> (those who can't replicate a table with auto-numbered
> column,etc).

--
Philippine Linux Users' Group (PLUG) Mailing List
[EMAIL PROTECTED] (#PLUG @ irc.free.net.ph)
Official Website: http://plug.linux.org.ph
Searchable Archives: http://marc.free.net.ph
.
To leave, go to http://lists.q-linux.com/mailman/listinfo/plug
.
Are you a Linux newbie? To join the newbie list, go to
http://lists.q-linux.com/mailman/listinfo/ph-linux-newbie

Reply via email to