Group: Forum Members
Posts: 386,
Visits: 2.1K
|
Hi Tim,
I've never used clusters but its my understanding that applications don't have to take clustering into account - it 'just works' with auto failover should one node fail. That being said I presume there's a bound to be quite a bit of new stuff on the admin side to create and monitor the cluster.
IMHO clustering is only useful in a "can't stop for anything" environment in which case you also need to be sure to cluster the box the apps run on.
We have a factory system in which when the computer dies the factory stops. What we did is accept that a stopage of 10 / 15 minutes was acceptable if it only happens once or twice a year and then made sure we had a near real-time copy of the database by using transactional replication which, in practice, never seems to have a latency more than 5 seconds. This is with a transaction rate of around 10,000 new records per shift. Then, if needed, we could restart on another machine using the backup database.
RAID disks are, of course, the first line a defence in securing the database.
Cheers, Peter
|