San Francisco, Calif.
Oracle believes the future of enterprise computing is the grid, and the enterprise software giant is staking its upcoming 10g
platform on it. Oracle used this year's OracleWorld Conference here to lay out its vision of grid computing, which CEO Larry Ellison summarized as "capacity on demand made up of low-cost parts," and to announce the first two products in the 10g
suite, Database 10g
and Application Server 10g
During his opening keynote address on Monday morning, Executive Vice President Chuck Phillips touted this year's conference as a milestone event for Oracle and the industry, saying the company was "ushering in grid computing." Ellison lent an historical perspective to that milestone during his keynote the following day. "Grid computing really is the first new concept in enterprise computing in 40 years," he said.
The concept of grid computing is on-demand sharing of computing resources among a group of servers (or grid) so that the idle processors of one server can be available to the others as needed. Oracle defines grid computing as pooling resources, virtualizing every layer of the stack, automating repetitive administrative tasks, and automatically load-balancing all servers based on customer-defined policies. Phillips explained that the concept is based on the utility model where users can request data and computing resources and get them on an as-needed basis.
In a complete, fully deployed 10g environment, a storage grid, a database grid, and an application server grid would all be monitored and managed from a single Grid Control interface, with each resource enabled with business intelligence that automates many of the tasks administrators currently perform (e.g., load balancing, provisioning, performance monitoring). "The beauty of our software," said Ellison, "is it appears as if it's all one big computer, to the application and to the administrator."
What's Wrong with My Architecture Now?
The problems that Oracle's grid push attempts to resolve are inefficient architectures made up of disconnected servers (they refer to them as islands of computation) and the costs required to scale these architectures when additional computer power is needed.
As Oracle sees it, many enterprises today have individual systems dedicated for each aspect of their businesses. Payroll and e-mail, for example, could each have their own servers and storage. These systems cannot share any of their capacity or resources because, as Phillips pointed out, servers are configured for maximum capacity at all times. Yet servers don't operate at peak capacity all the time and some use all their power only at certain times, like a payroll server at the first and fifteen of each month. Grid computing enables load balancing across systems to relieve the burden on "hot" servers and utilize "dark" servers.
Having dedicated servers also means that when one portion of the architecture needs more power the enterprise has to purchase additional servers, which can make scaling expensive. Oracle cites the low costs of building a grid as another compelling reason for adopting the concept. Phillips stressed how inexpensive the components needed for grid computing, blade servers ($4-5,000), modular storage (NAS and SAN), and interconnect technologies (Infiniband), are today. With only these components and consolidated servers, an enterprise can begin grid computing and the scale incrementally.
The emphasis on low-cost components seems an appeal to small and medium-sized companies that don't have the financial recourses or the existing financial investments in large mainframe data centers that large enterprises do.
Oracle would not give a specific release date for any 10g products. Database 10g and Application Server 10g are due the end of the calendar year.