Fetching directory, one moment please ...
Q. Why will Platform-as-a-Service be the killer cloud technology?
submitted by John Streeter
A. Because nobody should spend millions moving resource-wasting apps from data centre type A to data centre type B just the way they are, so they achieve nothing.
Self-service IT, elastic services and cloud bursting are very attractive prizes, which explains why Infrastructure-as-a-Service (IAAS) is the cloud technology that’s turning heads at the moment. But are those heads taking their eyes off the not-so-sexy, but very real problems and is this a seduction that will end in tears.
Possibly our biggest problem in the data centre is the classic Web Server - Application Server - Database Server architecture, with each component deployed on a dedicated server (physical or virtual) and each dedicated server not doing a lot. They’re called stovepipes a term that describes the vertical flue which isolates the application from everything else. Every stovepipe duplicates services every time. The outcome: enterprises running hundreds, sometimes thousands of application servers plugged to a similar number of database servers.
Stovepipes cost more than hardware and maintenance, they make every aspect of the the data centre really complicated. Things like the network, firewalls, storage, access portals, security and redundancy have to be worked and reworked to make each stovepipe functional. Complicated data centres need more staff and more processes just to keep running. High staff counts and rigid processes are antagonistic to creative, original thinking; which in turn guarantees ongoing proliferation of stovepipes.
IAAS has the potential to take stovepipe proliferation to a whole new level. When we virtualised the data centre, we were limited to migrating our existing stovepipes from physical to virtual machines. Now the hype-mongers are talking about self-service infrastructure - which to me reads like whole new classes of users spinning up new stovepipes on a whim. Didn't they create enough problems when ran up all those MS-Access applications?
Wouldn’t it be neater to aggregate our apps to a single Platform-as-a-Server (PAAS) infrastructure? The idea is to have a single application engine hosting everything. This is a scalable application engine: we can add and remove things like compute, storage and load balancing as needed. It is capable of running all our applications decoupled from specific server instances.
If we did that, we should see benefits like:
- We don’t have to guess our resource requirements up front on an app-by-app basis. We just deploy on the big platform and add more resources to the platform as the need arises.
- We can spin-up and tear down development and test instances of an application on-demand.
- We have a single place where our applications execute. We can locate that place in an appropriate security zone and manage access appropriately.
- We manage, develop and maintain our platform as a single entity. We have consistency and control – and we’re in a good position to really understand our environment.
- The application is only provisioned with the resources it needs, when it needs them.
How hard is this?
Our single platform is much more complex than any single isolated service. Logically we’re going to need fewer people to manage the beast, but they will need to be smarter people.
Doing this is different to what we’re already doing. It’s a task that requires the smartest technical people to make sure it delivers on vision and expectation.
Can it be done?
Various vendors such as Oracle and Microsoft offer private PAAS solutions and the market for public PAAS is mature with Microsoft, Oracle, Google and Rackspace being some of major players.
The good thing about a PAAS infrastructure is that applications can be migrated to the platform one-by-one and the resources from that application’s stovepipe and then be added to the platform if required.