|
Posted by Robert Klemme on 08/21/06 09:01
On 20.08.2006 19:59, boa wrote:
> Our situation right now is that the import job blocks all other traffic,
> rendering the system close to unusable and the customers unhappy. The
> idea is to move the import job related tables and indexes to separate
> drives to reduce disk seek time during that 1 crucial minute(or 33% of
> the time). This will hopefully improve overall throughput more than
> having everything on one array.
Thanks for the interesting read! I'd like to play a bit devil's
advocate to improve my understanding here - feel free to correct me.
Now, I figure that if you physically separate your bulk loaded tables
from other tables you're reducing IO bandwidth for this set of tables
and all others (compared to distributing all data across all physical
drives) by basically reserving parts of that bandwidth for specific
data. This means that an import job with this configuration will take
longer because it has less IO bandwidth available. If pages for the
import job tables displace pages of other tables from the cache other
operations will be slowed down more than for the distributed case and
they have to wait *longer* for the import job to finish because of the
reduced IO bandwidth. So it seems to me that Dan's approach (spread all
data files evenly over available disk storage) is superior in this
situation because then the whole IO bandwidth is available all the time
for all operations. What do you think?
Kind regards
robert
Navigation:
[Reply to this message]
|