X
Business

Oracle, BEA and the big virtual Java adventure

So there I was reading up on Mr Ellison’s announcement this afternoon that, “The addition of BEA products and technology will significantly enhance and extend Oracle’s Fusion middleware software suite,” and I thought a little deeper investigation into BEA was appropriate. BEA clearly had their house and their technology stack well in order for the “suits” at Oracle to scoop them up, so what better excuse for a little tech update and a spin around their developer resources.
Written by Adrian Bridgwater, Contributor

So there I was reading up on Mr Ellison’s announcement this afternoon that, “The addition of BEA products and technology will significantly enhance and extend Oracle’s Fusion middleware software suite,” and I thought a little deeper investigation into BEA was appropriate. BEA clearly had their house and their technology stack well in order for the “suits” at Oracle to scoop them up, so what better excuse for a little tech update and a spin around their developer resources.

Not one to wade through more than my fare share of white papers I was drawn to some interesting content from a chap (sorry, a BEA Technology Evangelist) called Martin Percival who writes that there is up to 90 per cent of unused server capacity in your typical enterprise and that server virtualisation can unlock some of that potential by allowing consolidation of multiple applications onto fewer physical servers.

According to Percival, for every $1.00 organisations spend on hardware every year, they spend $0.50 on maintaining it. Now I’m not sure if that is an accepted fact, a rough approximation or a product of the company’s PR machine, but it’s for sure that the old “need more power, buy another server” technique no longer holds water now that data centre space is at a premium. Perhaps costs are so high as we are all being so inefficient, IDC recently reported that there is $140 billion of unutilised server assets floating around in data centres. Do they mean stateside or worldwide? Either way, it’s like, a lot, right?

“In many data centres, one set of servers is assigned to each application; so it’s tough to move server resources from one application to another; and hard to provision new servers and processing power to meet the demands of applications,” says Percival. The trick of course is to bring in the virtualisation layer that contains a virtual machine monitor or ‘hypervisor’ to allocate hardware resources dynamically and transparently so that multiple operating systems can run concurrently on a single physical computer without even knowing it. OK OK bear with me, nothing too ground breaking there yet right? But the next bit on defining the hypervisor’s form and function is good.

“However, the hypervisor layer traditionally sits above the hardware’s operating system, adding a layer of overhead. Hypervisors are typically small in relation to a full operating system and consist of a collection of device drivers, memory management and the logic to time-share the various stacks above them onto the hardware. In effect, a hypervisor is a stripped-down operating system in its own right. With traditional hypervisors, a function call inside the Java application has to go through more layers to be executed on an actual CPU because it still has to negotiate its way through the existing operating system stack before even reaching the hypervisor. The hypervisor’s swapping in and out of each application’s state imposes additional overhead.”

But it doesn’t need to be this way says BEA. By adopting a two-pronged approach to virtualisation, organisations can optimise the operation of Java applications in virtualised environments - enabling Java applications to run directly on a hypervisor in a slim and efficient hypervisor-optimised software stack

This theory suggests that this two-pronged approach is bottom-up; by enabling higher resource utilisation and performance at the Java Virtual Machine (JVM) layer when deployed on virtualisation-enabled hardware; and top-down by enabling automated provisioning and resource management with adaptive control for enterprise Java applications. This approach is, so we’re told, able to improve utilisation by reducing the resources consumed by redundant and unused functionality in the stack and can allegedly enhance performance by removing layers between the application and the ‘bare metal’ of the server.

How much you buy into this concepts is up to you – but rest assured that the salesman will hit you with the following three buzzwords: flexibility, cost efficiency and environmental friendliness. There’s definitely a green angle to be had here. As I type this blog there‘s a radio news report going out saying that the UK is wasting huge energy resources by all of us leaving our mobile phones, iPODs and electric toothbrushes (or insert electrical appliance of choice) on constant charge even when they are powered up - so it’s not hard to see how legions of under-utilised servers wastes more than just a little of the national grid’s output.

Editorial standards