Grid Computing: Oracle Server Technologies and the Relational Paradigm
02.05.2010 by admin - 0 CommentPosted in Oracle
Critical to the concept of Grid computing is virtualization. This means that at all levels there is a layer of abstraction between what is requested and what is provided. End users ask for an application service and let the Grid work out which clustered J2EE application server can best provide it. Application servers ask for a database service and let the Grid work out from which RAC node the data can best be served. Within the Grid there is a mapping of possible services to available service providers, and there are algorithms for assigning the workload and resources appropriately. The result is that end users have neither the need nor the capacity to know from where their computing resources are actually being provided. The analogy often drawn is with delivery of domestic electricity: it is supplied on demand, and the home owner has no way of telling which power station is currently supplying him.
The Grid is not exclusive to Oracle. At the physical level, some operating system and hardware vendors are providing Grid-like capabilities. These include the ability to partition servers into virtual machines and dynamically add or remove CPU(s) and RAM from the virtual machines according to demand. This is conceptually similar to Oracle’s approach of dynamically assigning application server and database server resources to logical services. There is no reason why the two approaches cannot be combined. Both are working toward the same goal and can work together. The result should be an environment where adequate resources are always available on demand, without facing the issues of excess capacity at some times and under-performance at others. It should also be possible to design a Grid environment with no single point of failure, thus achieving the goal of 100 percent uptime that is being demanded by many users.
The SQL application developer need not know how the Grid has been implemented. The SQL will be invoked from an application server and executed by an instance against a database: the Grid will take care of making sure that at any moment pools of application servers and instances sized appropriately for the current workload are available.
