Databases Reference
In-Depth Information
Customer
Customer
Customer
Customer
Product
Product
Product
Product
Tr ansaction
Tr ansaction
Tr ansaction
Tr ansaction
Household
Household
Household
Household
Geography
Geography
Geography
Geography
Customer
Customer
Customer
Customer
Product
Product
Product
Product
Tr ansaction
Tr ansaction
Tr ansaction
Tr ansaction
Household
Household
Household
Household
Geography
Geography
Geography
Geography
MASTER CATALOG (Master Node)
FIGURE 9.2
Appliance data distribution.
The other performance benefits include:
●
Fault tolerance—if and when a node fails, a second node is available with the same data to process
other workloads, and you can stand up a backup node replica in a lot less time.
●
Workload management—data loading can be managed in different techniques across the products
while data querying and analytics are being executed concurrently. Due to advancements
in workload distribution beyond just setting parameters in the database, in the appliance
configuration, we can keep sustained performance during mixed-workload execution.
●
Another key workload management feature is the ability to dedicate resources across nodes
on-the-fly to accommodate completion of tasks. Smarter management of resources without
stopping the database or deleting job queues has created the flexibility for administrators to
manage workload shifting or adjusting to shifting workloads with simple techniques.
●
Node balancing—in a typical situation of adding more infrastructure resources, you do not get
to use 100% of the resources, as not all of them are available to you. In the case of the DWA,
each time you add a node, you get more self-contained infrastructure of memory, processor,
and storage, and all the 100% of the resources are available at that node's disposal. It is up to
the administrator to designate the node as a standby or replicate it to something specific. Once
the replicate option is selected, the data is automatically extended to the node and loaded. After
