Head to a conference like USENIX or LISA and you’ll find regularly scheduled hallway arguments over which systems management philosophy is the “right” one. Personal biases abound, but often the roots of these disagreements are simply differences in the perspectives formed through the combatants day-to-day responsibilities.
One guy will care more about provisioning networks and sees a server or a storage devices as just another object on his network. Another really doesn’t care about the network and lives his day-to-day at the kernel and process level of the OS. Yet another sees the OS as an almost static thing and is consumed with managing app frameworks, databases, and code releases.
Obviously, a “use the right tool for the right job” approach would be the sensible one, but the lack of definition around what the various “jobs” are pretty much kills those efforts. What you end up with is a lot of opinionated people insisting that their favorite tool is the right way of approaching any job.
The good news for all of us is that some clarity to these debates may be coming from an unlikely source… the mega-companies jockeying for position in SaaS and Cloud Computing.
Their contribution? Dividing up the online computing stack into well defined buckets.
For example, check out this presentation from Microsoft’s Chief Software Architect, Ray Ozzie (transcript and downloadable version of the slide deck). The presentation is from this past summer, but the direction of his thinking is very clear. Ozzie sees the stack in four distinct parts:
1. Global Foundation Services – Hardware, power, cooling, and connectivity.
2. Cloud Infrastructure Services – CPU time (think EC2), storage services (think S3), and network services like routing and firewalls.
3. Live Platform Services – All of the various distributed application and data services that your business or consumer applications rely on and share but don’t or can’t live in any one application.
4. Applications and Solutions – Complete business and consumer applications delivered using the SaaS model.
Of course, you can count on Google, Amazon, Yahoo (maybe), and probably SAP and Oracle weighing in with their own definitions. But the financially-backed handwriting is on the wall, the giants are going to force clear definitions one way or another. After all, they can’t sell a product you can’t name.
The significance? Well it may seem small at first glance, but it’s my prediction that the definitive naming of the layers that occupy our world and the glaring spotlight provided by the media coverage these giants receive (and the budgets that come with it) will be the catalyst for period of advancement in the systems management field that hasn’t been seen the 80’s.