Many recent science fiction novels deal with the concept of nanites - tiny bits of computers than can aggregate themselves to form new larger composites to assist their host. These concepts typically relate to making human cells self-healing, but they also have their sinister aspects, like the terminator, made out of liquid living metal in the last movie, who could reconfigure himself at will into any shape, and recover from any injury.

While this may seem like fantasy at the moment, there's a quiet revolution going on in the computing industry that makes this dream (or nightmare) one step closer to reality. Oddly enough, it's all a part of service-oriented architecture.

The underpinnings of any SOA are based on a service-oriented infrastructure, which basically creates callable, configurable services out of the lower-level components that make up an application infrastructure. Many of these services are familiar as system services - single sign-on, auditing, security management, reporting, etc. But what is fascinating in this world is the change that is taking place in the guts of the hardware.

Space in a data center is always at a premium, as is computing power. The natural tendency to try to pack more computing into the same space has been taken to its extreme conclusion with recent advances. Multi-core chips put two, four, or even more CPUs in the space previously occupied by a single CPU. Processing power is exploding like never before, and with cheaper memory and disk costs, the machines of today under the average user's desktop pack more processing power and capacity than many supercomputers of the previous decade.

All this would be trivial, except that with this rise we've also seen some very interesting services developed along with the silicon. Virtualization is a key part of the new wave of computing. With the ability to share resources down to as little as a 1/10th of a CPU, virtualization provides an amazing degree of flexibility.

Many developers and project leaders will tell you that the infrastructure costs for disaster recovery and testing are prohibitive. It is frequently impossible to fully fund multiple environments to support such requirements. With the advent of virtualization, the DR environment can be the testing environment under normal conditions, and then, at a time of emergency (or even just a request for increased capacity), can reconfigure itself quickly, ratcheting back or eliminating testing completely to become a full production environment, much like our terminator changing from one form to another.

The key to all of this is a service-based approach that allows for definitions of environments to be modified in response to changing environmental conditions. Theoretically no human intervention is even necessary - fail over at the router or load balancer will trigger a service call that can make the transition happen automatically. Similarly, with the return to service of a primary site, the fail back of the environments can reconfigure the disaster recovery site back to a testing environment.

Nor is this a binary condition - testing comes in different flavors and it's easy to imagine having multiple virtual environments to switch to - some with additional horsepower to support performance testing, others with less capacity to support functional testing instead. With automated testing tools, this can even take place in the complete absence of human intervention. At night the machines run themselves; in the morning, the humans analyze the results and plan the next evening's run.

Hundreds of other possibilities exist, from automating existing processing assignments, to reconfiguring networks, to adding additional disk capacity from a farm, the ability to morph and change the infrastructure is increasing rapidly.

Fortunately we're a long way away from sentient machines, and time travel still appears to be a fantasy with no real hope of ever happening. But the rise of SOA is certainly upon us.