In this month’s Research Spotlight, Open Research Europe presents an interview with Antonis Troumpoukis, a co-author of ‘A geospatial source selector for federated GeoSPARQL querying’.
This article was nominated by Dr. Fernando Aguilar, a Guest Advisor for the Cloud-Based Technologies Collection.
“The authors of this article take advantage of semantic and linked data technology to improve the interoperability of geodata services, which may make the data more reusable and findable, so I think it is a good way to improve the openness of data.”
– Dr. Fernando Aguilar
How would you summarize your Method Article?
Today, many data providers publish their datasets publicly on the internet. This presents an opportunity to develop federated query processors, also known as a federator. These are systems that combine several sources of publicly available data into a single virtual dataset, making it easy for users to access everything in one place.
In order to serve a query, the first problem a federator must solve is deciding which federated sources are relevant to which parts of the query. This problem is known as source selection, and the software that handles it as a source selector. A standard source selection technique is to annotate each source with information about the kind of data it contains and use it to remove sources with irrelevant data.
Geospatial data is data about objects that have a location on the surface of the Earth. This data is mainly managed by systems that have been optimized for the centralized processing of
However, geospatial data systems were not developed with the decentralized and open nature of sharing and discovering data in mind. Not much work has been done in federating geospatial sources, and no source selection methods that target geospatial data currently exist.
In our Method Article, we present a system that pre-selects data services relevant to a query using “summaries”. These summaries are shapes that encompass all the areas that the service has data about. By excluding sources with summaries of irrelevant geographic areas, our approach significantly enhances the efficiency of the querying process.
What inspired you to investigate this?
Over the past years, we have been involved in research on federated query processing and in the development of the Semagrow system. Semagrow is a versatile federated query processing system that presents a unified view of different data services, so that the application developers that use them do not need to be aware of what information is in which service.
A characteristic use case that inspired us to investigate the topic of our article was the case of food security in the ExtremeEarth project, within which the work presented here is conducted.
In this use case, crop type needs to be combined with nearby snowfall and snow storage information, since irrigation largely depends on snow storage and seasonal release of fresh water. Experimenting with this scenario in a federated data environment was the motivation behind some of our technical choices and backed our experimental setup with data and a query load.
What were the outcomes of your research?
In our Method Article, we introduced a new method for selecting data sources in a federated query processing context, specifically targeting federations of geospatial databases. Our method has proven to be correct for random federations and is configured in an easy-to-understand format for data service summaries.
We evaluated our method using data and queries from the agro-environmental domain, hence the use of ExtremeEarth as a characteristic use case. Even though we spend more time in source selection, the number of sources used is smaller. This results in more effective query processing because the federator has to issue fewer queries to the remote endpoints to evaluate the query.
We conducted an empirical comparison of three different types of summary shapes, ranging from simple shapes that contain much more area than the sources cover, to precise summaries that are overly complex and are more difficult to handle. Our experiments demonstrate their relative strengths and weaknesses and guide decisions regarding which summary should be preferred depending on the characteristics of different summarized datasets.
The results show that our method has a substantial positive impact on overall query processing time, and more precise selection makes the federation engine more sensible with the web resources it consumes.
Why is open research important to you and your co-authors?
Research in federated query processing and geospatial databases is fast-paced and it is essential for us to ensure that our research results are disseminated properly and can reach a wide audience. Publishing in journals that offer open access options makes research articles more discoverable online and freely accessible to everyone, including researchers in developing countries, individuals not affiliated with research institutions, and other members of the public.
Additionally, reproducing an experiment in the context of federated query processing is a challenging task since it is based on a complex setup that involves many datasets and several pieces of software. Making all underlying data and software publicly available offers other scientists the ability to reproduce our experiments and verify our experimental results. In addition, it allows our software components or our experimental setup to be reused in other contexts, leading to new research results, developments, and innovations.
What motivated you to publish with Open Research Europe?
Traditional scientific publishing tends to have several problems, such as publication delays and closed peer review. Moreover, in open access journals, the cost of publishing required by the authors can be very high.
We found Open Research Europe to be an attractive platform for scientific publishing since the platform offers rapid publication, an open peer review process that is fully accessible to everyone, and it is free for authors to publish. Moreover, we were interested in trying the peer review model of Open Research Europe, in which the authors are involved in the peer reviewer selection, helping to ensure the most appropriate reviewers are invited.
We were pleased with the peer review and publishing process, and we will definitely consider publishing with Open Research Europe again in the future.
What are your next steps?
In the future, we plan to continue our research on federated geospatial query processing in two main aspects.
Firstly, we will work towards finding proper geospatial summarization techniques for maximizing the performance of the source selection; there are many ways of balancing the trade-off between summary size and summary accuracy other than the three that we have already used in our current experiments.
Secondly, we are interested in exploring the deeper integration of geospatial summaries into all phases of federated query processing. As of right now, we have used them for identifying which of the federated sources refer to which parts of the input query, which is essentially only the first step of federated query processing.
Read the full Method Article on Open Research Europe to learn more about the method and findings.
