|
If you believe the Oracle marketing hype, then all applications would
benefit from RAC. But in truth, the ones that benefit the most are the
ones that need either high availability or higher scalability, or both.
If your application is running fine on a non-RAC implementation and you
don't need the higher availability that a clustered database would
provide, then I see no reason to make things much more complex by
implementing RAC.
It used to be that applications had to be segmented to get good
performance from RAC's predecessor, Oracle Parallel Server (OPS). OPS
had a problem when an application on one node needed data in another
node's buffer cache. The other node would have to write that data block
to disk first before the first node could access it. This was known as
"disk pinging." Towards the end of Oracle8i's life cycle, Oracle
introduced a technology called Cache Fusion. Cache Fusion is one of the
reasons that RAC is so successful for virtually any application. Cache
Fusion eliminated disk pinging by letting Buffer Caches in each node
move blocks back and forth across a high speed interconnect instead of
disk. Now that Cache Fusion is in place, there really isn't any
technological restrictions on applications that can successfully use
RAC.
|
