Advances in computing, communication, and software technologies have resulted in explosive growth of both information services and their underlying infrastructures. Information services are inherently complex, dynamic and heterogeneous, as are the large information infrastructures that support them. This is typified by the Internet, which comprises a large number of independent computing and communication resources. The amalgamation of the two results in management complexities that can easily break current computing paradigms and render both of them brittle, unmanageable and insecure. The shortage of skilled IT professionals in the current growing global economy further exacerbates the problem. To deal with such complexity, dynamism and uncertainty, researchers are taking design approaches modeled after biological systems. This research area is referred to as Autonomic Computing (AC).

IBM [1] coined the term Autonomics in 2001 and defined four main features to highlight it, namely, self-healing, self-protection, self-optimization and self-configuration. Since then other terms such as Organic Computing and Self-managed Computing have been applied to similar concepts, and the terms have been expanded to include features such as self-recovery, self-diagnosis, self-adaptation AC is inspired by the human autonomous nervous system, which handles human body complexities and uncertainties without requiring conscious effort. Clearly, the goal of AC research is to design computer infrastructure that manages itself with minimal human intervention. Of course, this is easier said than done. The first step in achieving the AC vision is to establish a detailed understanding of the underlying environment, including resource specifications, workload characterization, business processes and workflow, and service-level agreements. Only then can technologies such as provisioning, virtualization, monitoring, automation and feedback control be deployed to build AC features such as self-protection, self-healing and self-optimization.

Two major requirements must come together to build AC environments: (1) AC systems must be built of self-managing hardware and software components in which AC-related features are built into the architecture (not bolted-on); and (2) the environment must include a comprehensive management framework and architecture that facilitates building and deploying AC systems. Intel Corporation has been diligently working on both fronts, and is at the forefront of adding autonomic features to its products. Intel^® Active Management Technology (Intel^® AMT), is the first product incarnation of a framework and dedicated platform execution environment for AC. Intel has also been quite active in the AC research community, not only working on new AC research, but also developing vehicles to make it easier for the AC community to conduct related research in more realistic settings with a wide range of usage scenarios.

Although the AC research community has delivered good results so far in many AC-focused conferences and workshops, AC-related research is still in its infancy given the broad scope of the vision. One obvious reason is the correspondingly wide research scope, which can easily be grouped into two buckets:

1. AC Systems' Architecture, including (a), self-managing components such as processors, memory, network, storage, software blocks, and operating system; (b) guaranteeing that systems composed of self-managed components will always be autonomic, enforcing system-level policies and goals through establishing effective feedback control mechanisms principles such as stability, observability, and controllability; and (c) handshaking/interfacing among components within an AC system and among AC systems.
2. AC Systems' Intelligence, including fundamental sciences such as machine learning, heuristics, statistical computing, fuzzy logic, probabilistic reasoning and tracking/forecasting. Clearly, this bucket includes the widest scope and requires a great deal of attention.

Along with research work, there are many technical challenges such as:

1. AC Systems' prototyping, including advanced development of AC systems from subsystems with capabilities that allow us to emulate various autonomic scenarios
2. AC Support Ecosystems, including the software and hardware such as tools, orchestration software and lifecycle management needed to deploy AC systems in datacenters
3. AC standards for resource description, communication and messaging, etc.

In summary, we have made big strides in our efforts to achieve the AC vision, but there is still a great deal that needs to be done.

Reference
[1] Kephart, J. and Chess, D., "The Vision of Autonomic Computing," IEEE Computer, 36(1), 2003, pp. 41–50.