Figure 1: The Pervasive Semantic Grid Triangle.
David De Roure
Electronics and Computer Science
University of Southampton, UK
dder@ecs.soton.ac.uk
Pervasive Computing is the means by which the digital world of the Grid couples into our physical world. This paper presents a case for combined work on grid and pervasive computing, and for a semantic approach to both.
The paper is structured around the triangle of Figure 1, looking at each side in turn. In section 2 we consider the Grid and Pervasive axis - in each direction. We then move up: in section 3 we consider the Semantic Grid side and in section 4 we look at 'Semantic Pervasive'. Section 5 considers the whole vision and section 6 looks at how it is being applied in a case study. We finish in Section 7 with some closing remarks.
Figure 1: The Pervasive Semantic Grid Triangle.
Moore's Law tells us that if you keep the box the same size (the desktop PC, for example) then it will get increasingly powerful over time. If you want the same power then you can work with smaller and smaller devices - and more of them. We observe that, broadly, this gives us the world of the Grid and the world of pervasive computing, respectively.
Pervasive or ubiquitous computing is about devices everywhere; e.g. in everyday artefacts, in our clothes and surroundings, in the external environment. The term ubiquitous was adopted by Xerox through the work of Mark Weiser[1], who emphasised the 'calm' aspects, where the computing is everywhere but 'stays out of the way'. In Europe, pervasive computing is part of the Ambient Intelligence vision.
Some of our grid applications involve pervasive computing, including the coupling of sensors to the grid (such as in Sensor Networks, or a 'smart laboratory'), signal processing based on data from wearable devices, and the use of collaboration technologies. So the Grid does not necessarily present itself as a socket on the wall (like the electricity utility after which it is named) - it has a more pervasive coupling with the physical world, through what might be described as 'grid appliances'. If the Grid is the "back end" then the devices are the "front end" by which we interact with the Grid and by which data is collected. In other words, pervasive computing provides the manifestation of the Grid in the physical world.
Sometimes it is the Grid application that demands the pervasive computing and sometimes the pervasive computing that demands the Grid. In the former category would be the "grid-enabled" devices in a laboratory - the pieces of scientific equipment connected directly to the grid - and also the handheld devices with which users gather information and access results. Devices, such as novel interfaces, may also be deployed to support the collaborative environment, perhaps facilitating visualisation and annotation of data. In the category of pervasive computing that demands the Grid we have the sensor networks - as sensors and sensor arrays evolve, we can acquire data with higher temporal or spatial resolution, and this increasing bulk of (often realtime) data demands the computational power of the Grid. Meanwhile many pervasive deployments are currently small scale, due to small numbers of devices or small numbers of users, but will demand more grid processing as numbers scale up.
By way of example, both aspects come together in a project in which patients who have left hospital are monitored using wearable computing technology. Since the patient is mobile, we gather position and motion information (using devices such as accelerometers) to provide the necessary contextual information in which to interpret the physiological signals. The signal processing occurs on the Grid and medics are alerted - by pervasive computing - when the patients experience episodes that need attention.[2]
Both Grid and Pervasive computing are about large numbers of distributed processing elements. At an appropriate layer of abstraction, both grid and pervasive computing involve similar computer science challenges in distributed systems. These include service description, discovery and composition, issues of availability and mobility of resources, autonomic behaviour, and of course security, authentication and trust. Both need ease of dynamic assembly of components, and both rely on interoperability to achieve their goals. The peer-to-peer paradigm is relevant across the picture.
This abstracted view blurs the boundary between the Grid and the devices. For example, we can ask to what extent we can push the computation from the Grid back towards the devices? And to what extent can we achieve commonality in the middleware; e.g. how far can the grid services architecture be applied towards the devices side of the picture? We can envisage grid services relating to the sensors, be they pollution monitors, physiological signals, motion or location information.
There is a fundamental trade-off that is being explored here. On the one hand the computational power and storage capability of our portable devices increases as the technology improves. But at the same time we move into a world in which our devices can delegate storage and computation to the Grid, perhaps in real-time through greater connectivity and devices which are increasingly 'always on'. We suspect there is no one correct answer - the best solution for a given application can be clear, and in general we do need to explore both routes.
Perhaps most profoundly, Grid computing and pervasive computing are two visions of the future that really do seem to be upon us, and so surely they must be investigated together rather than in isolation.
In the Semantic Grid [3] we argue for the application of Semantic Web technologies to grid infrastructure and applications - the right hand edge of the triangle. The increasing extent of discussion about metadata within the Grid computing community is evidence that the Grid is increasingly concerned with issues of metadata and metadata management. The Semantic Web offers tools and techniques to address this.
The Semantic Grid 'holy grail' could be dynamic assembly of Grid components (software, services, data, knowledge) to meet the requirements of the Grid application and users. This requires everything to be machine-processable, i.e. metadata everywhere, with an agreed interpretation. We have argued that the full richness of the e-Science and Grid visions require this.
The Semantic Grid reality is that there are some things we can do now and some things which are still research topics. Clearly we should move towards a 'metadata-enabled' world as soon as we can, as this is a prerequisite for the sophistication of the full vision. The relevant technologies, such as the Resource Description Framework, are available for this. Meanwhile we can track the developments such as the OWL Web Ontology Language and the tools that are emerging to support it. This dual approach is adopted by the Semantic Grid Research Group in the Global Grid Forum.
Some applications of Semantic Web technologies in Grid applications are well aligned with the research and development activities in the Semantic Web community, most notably in areas such as bioinformatics where there is established use of ontologies. However, the application of Semantic Web technologies inside the grid infrastructure is less well developed. The emerging work on Semantic Web Services [4] is synergistic with this aspect of the Semantic Grid, reinforced by the adoption of a service-oriented approach in the Grid through the Open Grid Services Architecture [5] (OGSA).
There is increasing activity in the Semantic Grid arena, with an increasing number of research and development projects. The latest information on Semantic Grid activities can be found on the community portal www.semanticgrid.org.
We can also argue the left hand edge of the triangle - that the full richness of the pervasive vision also needs the Semantic Web technologies. Again this is about semantic interoperability. A key motivation for the semantic interoperability on the Grid is the need to assemble new Grid applications with ease (and preferably automatically), and surely we wish to do this with pervasive applications too. Essentially we have lots of distributed bits and pieces that need to work together to provide the requisite global behaviour, and we wish this to happen without manual intervention. This move towards a more autonomous approach might be achieved through the techniques of agent-based computing, and in the future we look towards self-organisation. This is the vision of autonomic computing.
Again we need service description, discovery and composition, and indeed research areas such as Semantic Web Services are being applied both to Grid and to Pervasive computing. However, semantic interoperability applies to content as well as services and it is clear that pervasive computing research and development projects do not always address these information and knowledge issues - in fact they pay less attention to such issues than Grid projects, which more often are motivated by information processing requirements.
The picture is improving with standards for representing e.g. sensor data, but we need the metadata too - for example, we need to achieve semantic annotation with context information at source. So metadata schema need to be established. To enjoy the promise of the Semantic Web we also need a means of creating and sharing URIs to get the added value of linking up disparate metadata via the objects it describes realizing the 'network effect'. Interestingly there's a whole new way of tying 'stuff' together here in the pervasive world because physical objects and physical space provide further ways of associating the disparate metadata.
We have suggested that pervasive computing needs the Semantic Web. We believe the converse is also true. One of the fundamental obstacles to realising the Semantic Web vision is that it requires metadata, and if we expect users to create that metadata then that in turn is a fundamental obstacle. Hence we need to automate metadata capture as far as possible: We need to take it out of the hands of the users and look instead to the pervasive computing devices to do the dull work. This is part of the total digitization discussed in Section 6.
As with Semantic Grid we will certainly find that we are pushing the Semantic Web technologies quite hard. For example, current solutions (such as triple stores) tend to favour a world of fairly static metadata - grid applications challenge this, and pervasive more so. There is much important work to be done on this edge of the triangle.
Let us consider all three - the full richness of the grid, pervasive and semantic visions. There are people who will argue that each of these futures will surely happen, and here we argue that there are clear reasons why they need to happen together. It follows that we need to be exploring that combined world - the whole triangle. Through combining grid and pervasive and semantic we see a comprehensive infrastructure for the vision of 'ambient intelligence'. It is the manifestation of the Semantic Grid in the physical world.
Exploring these three visions together requires working across at least three communities. With Semantic Grid we bridge the Web and Grid communities, for the "Semantic Pervasive Grid" we touch the devices and the next generation of the pervasive networking infrastructure too.
A number of projects are driving this forward, including the UK e-Science projects connected with the IRCs (Interdisciplinary Research Collaborations) and some under the Ambient Intelligence banner. As with Semantic Grid, we also need to put in place the community mechanisms to enable this all to happen. Organisations such as the Global Grid Forum (GGF) provide the necessary environment for agreement about those schemes that will provide interoperability. At the time of writing it is not clear what the equivalent body would be for 'semantic pervasive' computing?
The Semantic Web has become known for its focus on ontologies. However, there is a very fundamental part of Semantic Web philosophy that is all too often neglected and it is an extremely powerful part of the Pervasive-Semantic-Grid vision. This is the network effect that is achieved through everybody annotating the same URIs with their own metadata, such that the URIs then effectively link this metadata together, increasing its value. As mentioned in section 4, a similar grounding and linking can occur through spatial and temporal references, which is highly relevant in many pervasive applications.
For example, in the smart chemistry laboratory, we can use pervasive computing devices to capture live metadata as it is created at the laboratory bench or in the fume cupboard. As experiments occur we automatically record the data and metadata at source - and we use the devices to relieve the chemist of the burden of metadata creation. We refer to this as 'annotation at source'. This data then feeds into the scientific data processing - new data is derived from it, analysis occurs, as in Figure 2. All usage of the data is effectively an annotation upon it. If we can make sure everything is linked up through the shared URIs then scientists wishing to use these experimental results in the future can chase back to the source; i.e. the provenance is explicit. This is demonstrated in the World Wide Molecular Matrix [6] and we are investigating the pervasive and HCI aspects of this this in the Comb-e-Chem project through Smart Tea. [7]
Figure 2: Publication at Source in CombeChem.
In the Access Grid we also see annotation at source - people taking notes,
writing on a whiteboard, speaking, entering and leaving the room, etc.
For this to be useful in the Semantic Web sense we need this annotation
to create metadata with a known semantics.
In the CoAKTinG project [
The scholarly communications that are part of the scientific process are profoundly affected by the "publication at source" vision we describe here, in which experimental results can be made available with their associated data and the full information on the pipeline of processing that has occurred. In addition to data and papers, it is also possible to capture the entire digital discourse that occurs as part of the scientific process, and to incorporate this too into the densely interlinked outputs. This is a new approach to publication and it challenges the existing models of data repositories and scientific journals.
This paper has argued a case based on engineering - to build the systems we need to build, these are the technologies we need. The measure of success of the Semantic Grid, and of pervasive/ubiquitous computing, and of semantic pervasive etc, is that we have succeeded when nobody knows the system is there!
However, the real motivation for grid computing is the new science - and arts and humanities - that we can achieve with the power and scale of our computers. That is why we want to build the grid applications in the first place. Remember also the power of pervasive computing - such as new ways to augment our experience and improve our quality of life (and of course to assist those scientific discoveries). Those increasingly powerful devices coupled to the increasingly powerful grid fabric are increasingly capable of knowledge processing. Think about that world of knowledgeable devices.
Finally, we should challenge the "disappearing computer" and the aspiration to "seamlessness". Let us use these technologies to do new, interesting, creative and enjoyable things - to provide the e-researcher with new interfaces and artefacts to support their work.
This work is partially supported through the UK EPSRC e-Science programme grants Comb-e-Chem (GR/R67729/01), CoAKTinG (GR/R85143/01) and Grid-based Medical Devices for Everyday health (GR/R85877/01), the Equator (GR/R85877/01) and Advanced Knowledge Technologies (GR/N15764/01) Interdisciplinary Research Collaborations, and an an IBM Faculty Partnership Award. The talk on which this paper was originally based was given by the author at the IPv6 Global Summit in Madrid, June 2003.