New generation Intel-based servers are pushing
hard on the Oracle industry. Hardware vendors are calling out to
Oracle professionals, each promising lower Total Cost of Ownership (TCO),
faster performance, and easy scalability. With so many choices,
migrating onto a 64-bit platform can be a confusing proposition. In
general, the following are competing approaches to scaling the
Oracle data warehouse:
§
Vendors promise on-demand computing resources, lower TCO, and easy
scalability. Their huge servers offer savings from CPU and RAM
consolidation, far less human management costs, and seamless
allocation of resources.
§
Grid vendors offer solutions where server blades can be added to
Oracle as processing demand increases. While Grid computing offers
infinite scalability, no central point of failure, and the use of
fast cheap server blades, it does have the same in-the-box
parallelism that is found within a monolithic server. Unlike the
scale up approach, Oracle10g Grid computing is not automatic and
requires additional costs, additional training, as well as
sophisticated monitoring and management software.
Most savvy Oracle data warehouse shops practice
the scale up approach first. They only scale out when they reach
the processing limits of their server, which is a very rare
occurrence for today’s data warehouses. Many Oracle warehouse
professionals have learned that scaling out with Real Application
Clusters (RAC) and Grid is not an optimal solution. Instead, they
choose the scale up approach within a single server for many
compelling reasons:
§
Complex data warehouse queries need easy parallel query capability
and many on-board CPUs to maximize throughput. RAC nodes and Oracle
Grid server blades rarely have more than 4 CPUs.
§
Oracle clustering solutions are complex to configure and manage.
For the Oracle data warehouse, a large monolithic server provides
complete on demand resource allocation and scalability.
§
Oracle RAC licenses are expensive, and the DBA staff needs expensive
specialized training to master the complex inter-node
communications.
§
Unlike the scale out approach, a scale up Oracle data warehouse will
be instantly able to leverage new server resources without any
changes to the environment.
There is also a common misconception that using
a single server with scale up capabilities introduces a single point
of failure problem. In reality, hardware redundancy, on servers,
such as the Unisys ES7000 400 Series servers, offers further
protection against failure including redundant cooling, power, and
dual air conditioning with on-board power management, hot-pluggable
components, automated failure diagnosis and recovery, and proactive
failover mechanisms.
The scale up approach is the natural reaction to
the rampant distribution of Oracle systems onto small, independent
servers. This architecture saves money on hardware costs at the
expense of having to hire a huge system administration and DBA
staff. This is the appeal of consolidation: to avoid the high
overhead and expense of such server farms.
Due to the advances in the UNISYS server
technology, the concept of using a large SMP server for Oracle data
warehousing has become very popular. The scale up approach provides
on-demand resource allocation by sharing CPU and RAM between many
resources, requiring less maintenance and human resources to manage
fewer servers. More importantly, the scale up approach provides
optimal utilization of RAM and CPU resources and gives the warehouse
high availability through fault tolerant components. This approach
provides a high degree of scalability and flexibility for high
performance and also provides high availability and low cost of
ownership (TCO).
The scale out approach is designed for super
large Oracle databases that support many thousands of concurrent
users. Unless the system has a need to support more than 10,000
transactions per second, it is likely that the system will benefit
more from a scale up approach.
Conclusion
This chapter has made it clear that data
warehouse tuning is all about minimizing disk I/O. Oracle data
warehouses are disk intensive and require an architecture that helps
to keep the CPUs running at full capacity. As server resources
become like commodities, the savvy IT manager will choose the
platform that offers reliability, scalability and, above all, the
lowest total cost of ownership. In other words, go with the low TCO.
The main points of this chapter include:
§
What a Data Warehouse Needs
§
Oracle STAR Transformations and SQL
§
Why Oracle 10g is suited for the Data Warehouse
§
Methods for scaling the Oracle10g data warehouse
This chapter has shown the main tools and
techniques that can be used by the Oracle10g data warehouse
administrator for time-series warehouse tuning. With this
information on Data Warehouse Tuning with AWR, the next chapter will
delve into Oracle10g Tuning with the Oracle Enterprise Manager
(OEM).
