This content has been marked as final. Show 4 replies
In a quarter rack environment, if you lose a database node you will lose half of the database-side processing capacity. I'd suggest looking at how many resources your existing system is using, focusing on PGA size, buffer cache, and CPU usage. The PGA and cache advisors in particular can help determine what the impact of adding or reducing available memory might be.
Based on these results, your options might include:
- Determining that performance will continue to be acceptable with a node loss
- Use DB and I/O resource management to ensure that critical processing continues on a priority basis in a node-loss scenario (or do application-side changes to shed load)
- Upgrading to a half rack environment
And don't forget to consider other possible failure scenarios (storage server loss, network disruption, disk loss, etc)
I do not understand what you want to ask.
Are you saying the current configuration is 240G (of what? SGA? PGA? SGA+PGA? total server memory?) and that you are afraid that the two computing servers (database servers), which can be configured with 144GB each are not enough?
What is your current configuration? What are the requirements? The infiniband networking can relieve bottlenecks
If you want a technical valuable answer, you should provide enough details.
please, forget my considerations about actual memory. The question is refered to customer experience about memory consumption in database nodes when migrating to Exadata environment.
The question was pretty general so I'll take a stab at a very general answer. Memory usage on Exadata compute nodes (for buffer cache) tends to be less than on other platforms. The reason is that we generally don't have to play the index everything to death game to get the performance we need. (i.e. the optimizations provided by the storage software means we don't have to try to cache everything on the DB servers). When we do consolidation projects we generally measure existing memory usage and then apply a reduction factor to the buffer cache portion (we do to same thing to the cpu requirements due to the offloading capabilities of Exadata by the way). The actual reduction varies and that's where art comes in. Some types of applications make better use of the Exadata optimizations than others.