EXPERT RESPONSE
You have asked one of those rare questions that, in my opinion, has a clear-cut answer without any gray area.
In my mind, there is absolutely no question that shared storage is a far superior method for building a cluster. Why?
- Shared data is simpler. To keep two independent data stores consistent, you need replication. To keep one independent data store consistent, you don't need replication.
- Shared data is cheaper. You don't need a fat-piped network between the systems for replication and you don't need to concern yourself with replication software and its associated expenses.
- Shared data is more reliable. With asynchronous replication, it's always possible, even over short distances, that some of the replicated data will not have made it from the primary to the takeover system after an outage.
- Shared data is faster. If you replicate with synchronous replication, even over short distances, there may be some degradation of performance on the primary system. Not always, of course, but sometimes.
- Shared data is easier. If system A fails and system B takes over (depending on your method of replication and the nature of the outage), after A returns to service you'll need to send updates from B back to A before A is ready to come online. And, more significantly, if there is a power outage that affects both A and B while the data on the two systems are inconsistent, there is no way to know which system should come online as the primary. If the wrong one (A) comes on line data that had been written to B will be lost and it will be difficult, if not impossible, to reconcile it.
Honestly, I have never seen a persuasive argument in favor of replication between two clusters.
Hope this helps. Thanks for writing.
Evan L. Marcus
Editor's note: Do you agree with this expert's response? If you have more to share, post it in one of our discussion forums.
|
