More details about the new Oracle 10G and Application Server 10G are becoming available the closer we get to Oracleworld, and it's becoming clear that the 'grid' element of the new products is mainly an enhancement to the existing Real Application Clusters feature of 9i, with improvments to the way SQL statement are distributed amongst servers in the cluster, coupled with improved distribution of transactions and applications within an Application Server cluster.
According to this article at eWeek, the new components within the 10G database are aimed at increasing scalablility and speed, improving messaging, and making data storage more self-managing. Oracle's vision is for the 10G database to be the centre of the computing universe, co-ordinating processing and allowing users to push, pull and transact information from a grid of servers. Beta testers for have reported that new elements within 10G that support grid computing include improved XML handling enhanced Web services APIs and 8-exabyte file support, and of course improvements to the existing Real Application Clusters feature.
Other new features reported by the article include the ability to export definitions and other metadata through procedure calls, enable a feature called 'Transportable Data Spaces', an improvement to transportable tablespaces that's been enhanced so that data can be moved across servers from different companies. The Flashback feature introduced in Oracle 9i is being improved, with the ability to selectively recover individual SQL statements, whilst 'Big File Tablespaces', presumably the ability to create tablespaces with datafiles bigger 2Gb, is part of the self-storage management features that's been mentioned as central to grid computing.
The next version of iAS, originally known at 9.0.4 and due for release in the first half of 2003, is now going to be Application Server 10G, with the grid features mostly based around the distribution of processing around multiple servers to wherever resources are most available. This article at informationweek talks about how Application Server 10G differs from the academic model of grid computing, where a central server scavenges for unused processing cycles from computers on the network, as this may not be predicatable enough for businesses looking to run financials applications, close off books at the end of the month, or run payroll to a set deadline.
The point of Application Server 10G seems to be to spread the workload amongst a cluster of servers (or farm, as it's know with 9iAS), in the manner of a grid but with the luxury of knowing that there'll always be a certain amount of processing power available at each node, as it's your grid (rather than being public) and you can ensure processing power is available. Again, like the Database 10G, the clustering ability is based on that found in the 9i version (in this case, 9iAS), and, whilst a 9ias cluster could house different applications on different nodes in the cluster, 10G Application Server will be able to spread the workload of a single application across multiple nodes in the grid.
Central to all of this appears to be enhanced automation and monitoring tools, with the 10G Application Server able to provision a grid, identify and assemble its consituent parts for a given workload, assign tasks at particular times, and dynamically manage the workload, together with improvements to Enterprise Manager which centre around the ability to manage multiple database and application servers from a single web interface.
Now that details of Oracle's implementation of computing grids is starting to become clear, competitors such as IBM are questioning whether this extension to RAC is really what grid computing is all about. In this article at internetnews.com, an IBM spokesman said "Oracle is making a lot of noise about grid with 10g, but the way Oracle defines grid seems very different from IBM's definition; from what I can tell they are more or less rebranding RAC as 10g...meaning you still have to move everything into an Oracle database."
According to the article, Oracle's implementation of grids if markedly different from the approach of IBM, which provides a virtual view of information stored anywhere on a computing grid and doesn't require a database cluster to function in a grid environment. However, IDC's Carl Olofson thinks it's natural that IBM would think that, having effectively had the grid initiative taken from them, and that Oracle has RAC to something of a "closed grid" system where the servers are controlled by Oracle, allocated services as and when they are needed. "There have been serious improvements," Olofson promised. "They have done some clever things."
11:54:57 AM 
|
|
