Grid Computing Oracle Style
By: Rob Reilly
Thursday, February 6, 2003 09:58:20 AM EST
URL: http://www.linuxplanet.com/linuxplanet/reports/4661/1/
Oracle, Linux, and The Grid
Oracle has been at the forefront of database technology for a long time. The have also embraced the Linux environment in a big way. Now they've gone a step further, developing industrial strength database applications that run on clusters of commodity blade servers or grids. As we will see, they have produced bulletproof, scalable systems that should warm the heart of any CIO.
Benny Souder is Vice President of Oracle Distributed Database Development. In a recent interview, he talked about Oracle's 9i product, the Linux-based grid platform, and ways that the combination benefit the customers.
"Oracle is focused on all aspects of data management. Specifically how to make data available and how Oracle 9i can help clients," Souder said.
"The main idea behind a grid is that it is thought of as a utility. Much like the electrical utility grid. The client and user don't necessarily know where the computers or data reside or how it got there, just that it is available with high reliability. Power company customers don't necessarily know where their electricity comes from. Many tasks go on behind the scene (in both industries) to make the resources available, any time, any where," Souder continued. He called the concept "virtualization."
Virtualization and Provisioning
The mechanics behind "virtualization" is called "provisioning." There are basically two kinds of provisioning in the context of Oracle databases.
- CPU provisioning
- Data provisioning
CPU Provisioning
Using the Oracle Grid database, CPUs can be added and subtracted as needed, even when an application is running! Why would you want to do that? Suppose you were a Fortune 100 company that had to run regular monthly financial general ledger reports. Toward the time when general ledger reports needed to be generated, more CPUs could be added to handle the increased computational load during that time. When load (regular monthly transaction processing) was low again, those CPU "cycles" could be used in some other area of the business.
Data Provisioning
Provisioning data allows transportable databases, even portions of a database or individual tables. Tables can be detached from one database and re-attached to another database. Looking again at the financial DB example, some data might be needed to enhance the monthly general ledger report. The data could be connected to the general ledger function until the reports are completed, then it could go back to original database for normal updating, editing, etc.
Oracle handles this by manipulating database "meta" data, instead of shutting down DB, moving table and then restarting application. The moving of data and tables is done logical, instead of physically moving the files physically to a new location on the disks.
A companion product, called Oracle Streams, allows the detached data to be updated according to user's criteria and rules. This makes the data, in effect, location independent.
The Oracle database engine on top of Linux figures out how to "provision" CPUs and data seamlessly within the application, even while the database is running and being accessed.
RAC One Up for Oracle
An Oracle database running on a grid is known as a RAC or Real Application Cluster. A RAC runs great on commodity hardware clusters.
The normal hardware for RACs are blade servers. According to Souder, blade servers and commodity hardware compliment Linux and Linux compliment blade servers and commodity hardware. The Oracle Grid lends itself well to this environment, because it is scalable and the hardware can be swapped in or out as needed.
Blades are made by HP, Sun, Dell, IBM and other vendors. Other companies have database applications that they say run on a certain cluster. Oracle ups the ante and certifies that an Oracle Grid will run on a system. Currently, the Oracle Grid is certified on 32 node Linux/Intel clusters and 256 node clusters running AIX. Other cluster configurations will be certified as customer demand dictates.
"Oracle runs real applications, and is certified to do so. The key point is that other vendors cluster database have never been certified to run real applications. The other vendors just use their cluster database for benchmarks. We will certify the database running on more nodes as customers need more nodes--there is really no architectural limit to the number of nodes we can support," Souder said.
Other companies have so called "distributed" applications, Oracle has the code and ability to back up claims. As an example, the FAA was looking for high availability for their systems. But, it had to be affordable. The Oracle Grid was good fit because the database application didn't have to be shut down to service hardware or when a hardware fail-over occurred.
Another feature of the Oracle Grid Database is that is an application can be moved from one platform to another. It doesn't matter if an application was created on AIX (IBM mainframe) and client wants to move it to Linux (on an Intel based Blade cluster). The Application can simply be moved from one platform to another without any code changes.
Customers can also manage their Oracle Grid Database either from a GUI or the command line. The GUI is good for everyday simple tweaks. The command line is a favorite preferred by DBA's because it allows the scripting of tasks. If you can't decide which to use, the GUI has a record mode that can create a script of tasks, for you. You can then command line it from there.
Oracle Eats Its Own Lunch
How's this for trust in the technology... Oracle uses the Oracle Grid to design and write the Oracle database code itself. There are about 8 million lines of code that make up the Oracle 9i database. The Grid that manages all that code runs on 1300 blade server nodes. Cost savings of 90% were realized using this hardware setup as opposed to other solutions. Think of how credible the Oracle staff should be by depending on their own tools.
Support for the Oracle Grid is via vendor supplied assistance and the Oracle Tech Network. The Tech Network is like the Linux development community in that forums and mail lists are used for feedback and collaboration. Go to the Oracle Tech Network for information and downloads on Oracle's product lines.
Souder concluded, "Oracle is proud to offer products that are fully tested and ready to integrate into production."
Rob Reilly (aka: "Dr. Torque") is a senior technology consultant, whose work includes Linux, business systems integration, innovation training and occasional hot rodding excursions. He frequently writes and speaks about these and other topics. He has 16 years experience in the high technology, manufacturing and the utilities industries. He is always 'on-the-lookout' for stories and projects that focus on Linux, business and the cutting edge. Send him a note or visit his web site at http://home.cfl.rr.com/rreilly.