| The Neuromancer project is an investigation into large-scale distributed computing of the sort that will be brought about by the wide deployment of sensors and actuators. The deployment of such devices will change the edge of the network from a set of client devices (such as cell phones and personal digital assistants) to a set of services, feeding information into the rest of the network. The artifact that we are building to further this investigation is an infrastructure for wide-scale medical sensing. The infrastructure should enable the collection and use of information gathered from medical sensors deployed over the scope of a medical practice, a hospital, a metropolitan area, or an entire population. Further, the driving assumption is that these sensors will be deployed not just when there is a specific need, but all the time, allowing constant in situ data gathering of basic medical information. Further, this information should be kept for (at least) the lifetime of the patient, meaning that the system itself will need to exist for a time scale not usually thought of in computer systems. The scale of such a system, in terms of number of data sources, size of deployment, and estimated lifetime of the overall system, force us to address a number of major research questions. Among these are:
- Federation: given the structure of medical science, instances of the system must be possible to deploy at a local level yet be incorporated into other instances to form cooperative federations;
- Distributed Data Analysis: Because of the size of the overall information gathered, it will be impossible (even if it were desired) to store all of the gathered information in a single place. It will be necessary to find ways of finding, analyzing, and manipulating the information that is stored throughout the federation of systems;
- Extensibility: The information gathered by the system will have a lifetime similar to that of the patient. During that period of time, new sensors will be created that will need to be incorporated into the system.
- Mobility: The system has a number of notions of mobility. The people wearing the sensors will be mobile, requiring that we find a way of insuring that the information gathered about them is not lost even when they are out of touch with any network. However, the information and services that make up the overall system will also be mobile, since over the lifetime of the patient the machinery on which the system runs (and therefore on which the data and services reside) will be replaced, requiring that the data and services migrate from the old to the new.
- Versioning: In a system that is meant to last for 80 or more years, there will be times when new versions of sensors, data, or services will be introduced into the system. Given the physical scale and distributed nature of the system, it will be impossible to do simultaneous update of all the system participants. So the system will need to have a notion of versions built in, allowing old versions of services, data, or sensors to co-exist with new versions.
In addition, we will need to define a distributed platform against which services and data repositories can be written. The platform must insure uninterrupted running of the overall system while allowing the individual pieces of the system to be upgraded, replaced, and repaired. While the research is grounded in the particular application of large-scale medical sensing, we believe that any results will be applicable to a wide variety of other networking scenarios, including automotive or aeronautic telemetry and general large-scale, long-lasting distributed systems. |
