Sponsored by: Oracle Corporation
Carl W. Olofson
The IT world is in the midst of a sea change in how computing resources are
managed. The previously normative models based on program code assigned to
fixed assets are being replaced by models based on flexible code deployment and
resource assignment. A new model has emerged based on a shared infrastructure,
where the underlying physical hardware is abstracted from the application and logical
resources built by aggregating or disaggregating the physical resources that are used
for application deployment. This shared infrastructure, also known as grid computing,
provides many benefits to today’s IT operations. The benefits of grid computing in
general, and Oracle’s exploitation of it in Oracle Database 11g, are:
- Ending the paralysis that results from having computer and storage systems locked down on specific applications, databases, and files
- Enabling operations and administrative staff to flexibly assign compute and storage resources as needed, enabling them to avoid locked-down and overprovisioned resources
- Providing a means for easily assigning and tuning resources to meet the business’ service levels
- Reducing cost and enhancing business agility by more effective utilization of resources using commodity hardware
IN THIS WHITE PAPER
This paper discusses the evolution of the datacenter toward the adoption of grid
computing and details the benefits derived from this development. It considers the
implications for database systems and details the requirements of a database for
successful deployment in a grid computing context. It then reviews the features of
Oracle Database 11g and shows how that product meets those requirements.
Most datacenters today feature a range of computer and storage systems, all
dedicated to specific workloads or combinations of workloads, usually carefully
segregated by application and database. Typically, the computer and storage
systems are even named after the database or application they serve. Switching them
around to achieve greater efficiency is simply not done; instead, they tend to be
overprovisioned to ensure sufficient capacity for peak periods and for growth within
If storage utilization grows faster than the expected outlay, a complicated and
expensive project to reallocate and expand storage results. If computing power
becomes insufficient to handle the load, a “forklift upgrade,” consisting of replacing
the computer systems in question, is usually required. These actions are generally
accompanied by substantial staff efforts to load and configure the new systems and to
convert or migrate the existing workload to the new systems.
One way to reduce the frequency with which this occurs is to allocate much more
system capacity to given applications or databases, allowing them to “grow into” the
underutilized capacity over some manageable period of time. In the interim, of course,
a good deal of system and storage capacity goes unused.
What Is the Grid?
The grid represents an alternative to this expensive model of fixed IT assets. It
derives its name from its resemblance to a power grid, which enables power
producers (such as generating stations) to be added or removed as needed and
likewise power consumers to be added and removed, without complicated
configuration changes. An IT grid enables server and storage resources to be added
to or removed from the system without requiring complicated configuration changes,
thereby enabling flexible deployment that avoids the problems inherent in the fixed
Definition of Grid Computing
Grid computing is a term that has been applied to various architectures designed to
deliver the benefits of an IT grid. It is an approach to computing that detaches the
software functionality from the specifics of hardware deployment by blending system
and storage resources into a continuum of resources that can be allocated to, and
deallocated from, a particular function or functional locus; in this case, a database.
Simply put, it enables administrators to assign computing tasks to computing
resources, and it assigns data to storage resources in a way that enables such
resources to be easily added or removed or tasks and data to be moved as needed.
Various vendors have taken different approaches to delivering grid computing; these
approaches tend to vary based on the type of workload involved and the
requirements of that workload. They include infrastructure-based distributed
components for high-throughput compute grids, redundant application server
deployment for application grids, and various approaches to providing scalable yet
flexible storage management through storage grids.In the case of database workloads, grid computing contrasts with the classic model
that involves dedicated servers associated with dedicated storage in that the servers
and storage are fluid; that is, they can be assigned, added, and reassigned as
necessary without upsetting the overall topology of the database server environment.
The key benefits of grid computing come in the form of resource flexibility, scalability,
and optimization of operations through parallel processing. These benefits are
expressed through an architecture that gives users the following capabilities:
- To avoid unnecessary hardware, power, and staffing costs of overprovisioning IT systems, commonly done to avoid capacity upgrades.
- When capacity upgrades are necessary, to scale incrementally by adding (or in some cases, redeploying) system and storage resources without expensive “forklift upgrades” or time-consuming and error-prone upgrade procedures.
- To ensure continuous availability through the provisioning of redundant resources, ensuring automatic failover when necessary.
- To increase transaction throughput through parallelization of tasks.
All these benefits combine to enable better business agility in responding to changes
in load or business priorities.