Database Reference
In-Depth Information
(=per server) is reached. Now if for some reason an execution plan changes and response times for queries increase
it is likely that the application servers will start spawning new sessions to compensate. But instead of 10 times 10
sessions the database now has to deal with potentially 10 times 100 sessions which is a much higher workload. And
many of these sessions will actually be busy executing code, further reducing latency and so own. If the DBAs cannot
quickly rectify this problem a downward spiral that can even lead to the complete overloading of the database server
as shown in Figure 2-6 .
Figure 2-6. Structural view on the DRCP exposure for JDBC applications
When using a DRCP on the database host this effect can potentially limit the impact of spikes of activity on the
underlying infrastructure. As you can see from Figure 2-6 , the middle-tier remains the owner of their own connection
pools. In addition, each RAC node has its own DRCP to which the connection pools connect. Just as with “plain”
DRCP, the effect of connection storms can potentially be better contained.
The correct solution to the problem however would be to reduce the number of processes working on the
database and to set the minimum pool size equal to the maximum pool size to avoid unpleasant surprises.
Copy-on-Write for Cloning Databases
One of the problems frequently encountered for the database administrator is the exponential growth in storage
requirements. Imagine for a moment that one of your operational databases is 10 TB in size. If the business requires
five different test environments for different streams of work than those 10 TB quickly become 60 TB which can no
longer be considered a small footprint. The fact that storage is considered cheap can no longer be considered true
for everyone. Enterprise class storage is expensive, especially when you need lots. Some companies have long-term
agreements with the large storage providers, rendering them less flexible when it comes to offering discounts to
internal customers. The high price tag for additional storage has made requisition of additional terabytes or even
gigabytes a lengthy process with a lot of red tape around it.
Thankfully there is a way to reduce the storage requirement for test environments, and it is built in with the
Oracle database. The benefit of the new technology increases with every additional copy of the live environment.
The underlying technology is called copy-on-write. Instead of requiring full copies of the database for each test
 
Search WWH ::




Custom Search