Emerging Frontiers in Miniature Computer and Embedded Systems

by Angelos Bilas and Evangelos Markatos

Although over the last several years computer systems research has focussed on building general-purpose PCs, current emerging technologies and societal needs are shifting the focus towards ambient intelligent systems that monitor, adapt, and react to humans and objects.

Over the last twenty years, computer systems and architecture research has addressed the performance of general-purpose microprocessors and desktop computers, with impressive results. Processor performance has been doubling every eighteen months, and today's general-purpose desktop systems are used in most applications to address single-user needs or as components of much larger computing and storage platforms. However, there is now an evident need to turn the focus of systems research from the above targets to new directions, moving from general-purpose to application-specific systems and from performance to reliability, availability, security, autonomy, scalability, and manageability.

Current research at ICS-FORTH, which is partly funded by the 'Excellence' project initiative of the General Secretariat for Research and Technology of the Greek Ministry of Development, builds on existing expertise and addresses fundamental problems. These include:

miniaturisation of computer and embedded systems
interconnection of such systems
autonomy and coordination of very large numbers of such systems.

It is expected that such autonomous, networked, miniature computer and embedded systems will find application in many new areas, eg diagnosis of epidemic diseases, forest and crop protection, human safety and monitoring of fine-grain parameters in the human environment. Overall, miniature computer and embedded systems are predicted to be the most effective carrier of ambient intelligence. Such systems can both perform intelligent functions and be placed or integrated in locations where traditional systems cannot fit, due to form factor, autonomy, and cost restrictions. Moreover, they will have the ability to communicate with each other, providing a means of transferring, monitoring and controlling information everywhere in a ubiquitous and transparent manner. These directions in computer systems research at ICS-FORTH are discussed in more detail below.

Going small and going in numbers
In this regard, our research is searching for ways to build very small (say, coin-size) and flexible smart devices and embedded systems. Looking at today's state-of-the-art computing systems, two of the main factors restricting miniaturisation are the requirements on versatile I/O interfaces and rich runtime systems. The first is a result of various standard interfaces that need to be supported, whereas the second is a consequence of requirements for general-purpose functionality. Our goal is to re-think both of these issues. Firstly, we are interested in designing interfaces that can be used to connect components within very small systems. Of principal interest here is connectivity to non-logic-based devices (MEMS). Second, we would like to explore architectural support for novel runtime systems. Today's runtime systems provide flexibility that is not needed in application-specific devices, leading to high requirements in compute cycles. Our goal is to investigate how such devices should be designed to best support novel runtime systems that exhibit the necessary characteristics.

Connectivity is the only way in and the only way out
This focus area examines how such systems should be interconnected among themselves and also to the outside world. Building interconnects for large numbers of miniature devices introduces constraints that are not present when building traditional interconnects. The physical medium used is constrained not only by size and speed but also by power consumption, and may well limit the distance at which such devices can transmit and receive data. Moreover, given the requirements of future miniature embedded systems, how their protocol stack should look is unclear. Traditional protocol stacks are inadequate due to the size, complexity, and the requirements they need to fulfil. In addition, security may be jeopardised since, on the one hand, the network is the only means of gaining access to such systems, while on the other hand, a full access and management API exposed over the network makes these systems vulnerable to attacks.

All for one and one for all
Miniature systems are unique compared to existing computers because they are able to contribute to new application domains simply by being replicated in large numbers and by collaborating on solving a specific problem. However, this aspect of collaboration, which assumes a high level of autonomy, is new territory that needs to be explored. Previous work in distributed systems has addressed issues in systems that are more closely supervised, more independent in performing various tasks, and rather loosely coupled. The runtime system is a key component that implements most of the functionality relating to achieving collaboration and providing autonomy, but also that relating to tolerating and recovering from faults, providing reconfigurability, security, power management etc. Given that these systems will be less capable than modern computers, to a large extent the runtime system needs to be redesigned.

In summary, computer systems research is shifting its focus to new and exciting directions. We believe that among the most promising new aspects of computer systems research is the miniaturisation of networked computer and embedded systems and the implications of using large numbers of such devices. Large areas of uncharted territory remain in these fields for computer architecture and systems research.

Link:
http://archvlsi.ics.forth.gr

Please contact:
Angelos Bilas and Evangelos Markatos, ICS-FORTH
E-mail: {bilas, markatos}@ics.forth.gr