|
These days clustering is becoming synonymous to high availability. As an enterprise architect I have to shake my head in disbelief. IMHO, clustering can be a component of HA but is far from being HA in itself. I'm always used to approaching HA as a solution involving multiple components from replication to mirroring to DR. Does anyone else feel the same on this topic? Or am i just an grumpy old architect!
| |||
|
feedback
|
|
No, you are obviously right. HA is a goal, and clustering is one technology that helps you reach it. Clustering will not get you HA if for example everything goes via a single switch... | |||
|
feedback
|
|
Depends on your definitions I've always felt - I've always thought integrity of data is key component of HA and clustering doesn't give you this given the shared storage it employs. Any cluster needs to be backed up by a robust backup strategy that would include a blend of mirrors, replication and off-line backups as appropriate to the requirements of the system you're protecting. I've seen first hand examples of a cluster breaking down and there being no other plans in place. | |||
|
feedback
|
|
Clustering IMO is just like you thought, just a part of HA. It adds redundancy to prevent single system failure. However, just like any good HA environment, you should remove any single point of failure. Just because the servers are clustered, that doesn't assume multi-pathing to your storage, redundant switches in your network, or even redundant power circuits and PDUs in your rack. Replication is huge too, site failure can kill even the best HA solution unless data goes offsite (tape or otherwise). | |||
|
feedback
|
