Monitoring and evaluating the quality of Web Map Service resources for optimizing map composition over the internet to support decision making

Abstract

Over the past 10 years, there have been great advances in the interoperability technologies in geographic information science. More than 10,000 map layers are available online today through Open Geospatial Consortium (OGC) specified interfaces, such as Web Map Service (WMS), Web Feature Service (WFS), and Web Coverage Service (WCS). These map layers are persistently serving the geospatial communities; however, our empirical study found that their potential value has not been fully exploited. Frequently, a targeted map cannot be composed because some published map servers are unavailable. This problem becomes more serious when a map is composed of several layers from different servers. These services are geographically distributed and maintained by various hosts; therefore, simply waiting for service improvement on the host side cannot solve this problem. In this paper, we proposed a new approach and developed a mechanism that allows clients to select the best map layers at run-time. The selection is based on the results of continuous monitoring and evaluation of the quality of WMSs. Based on Service Oriented Architecture (SOA), this approach includes quality monitoring and evaluation modules. Quality factors are taken into account during the process of registration, search, and bind. The OGC capability document is extended to include WMS quality information. Three prototype systems were developed in this study to demonstrate: (a) how WMS layers are monitored and evaluated, (b) how the subjective evaluation of WMS quality by a user is collected, and (c) how this can be a feasible method to fuse WMS resources suitable for decision making.

Introduction

Over the past 10 years, there have been great advancements in the interoperability technologies of geospatial sciences (Yang and Raskin, 2009, Riedel et al., 2009). Interoperability is the ability of two or more systems or components to exchange information and use this information that has been exchanged (Geraci, 1990). Interoperable systems can communicate, execute programs, or transfer data among various functional units in a manner that allows the user to have little or no knowledge of the unique characteristics of those units (ISO, 2005a). Whereas the interaction between two systems may rely on private protocol (Wu et al., 2005), the interoperability among arbitrary systems is more desirable and must conform to public standards (Gadre et al., 1990). Based on standard protocols, Service Oriented Architecture (SOA) has become a major framework for web-based geospatial applications. Foster (2005) even proposed a concept of Service-Oriented Science to suggest that the future of scientific development will be based on this architecture. In this way, tools that were formerly accessible only to the specialists will be available to all. Previous manual data processing and analyzing tasks can be automated through service chaining.

Driven by applications, interoperability between geographic information systems is particularly advantageous. The high cost of geographic data acquisition and the heterogeneous format of different conceptual models create an instant demand for interoperability (Zhang et al., 2007). Interoperability is widely recognized as a new paradigm for integrating distributed and heterogeneous computer systems to facilitate a more efficient use of geographic information resources (Harvey et al., 1999). From inception, industry and academia have attempted to design a widely acceptable conceptual model to unify geographic data interoperability. Open Geospatial Consortium (OGC) designed and published abstract and implementation models for manipulating geographic data at the feature level (Kottman and Reed, 2009). The simple feature implementation specifications (OGC, 1999a, OGC, 1999b, OGC, 1999c) have gained wide support from most commercial GIS software developers. For external geographic information exchange, Geography Markup Language (GML) has been developed to act as an open interchange format by encoding a spectrum of geographic application objects and the associated properties (Cox et al., 2002).

Theoretically, these interfaces defined by simple feature specifications and GML are effective and easily integrated into heterogeneous GIS platforms in a local area network (LAN). However, the fast growing popularity of the Internet and loosely coupled SOA applications, such as web mapping and decision support systems, call for more interoperability in the wide-area network (WAN). There is no doubt that the Internet, which has been considered the most revolutionary technology of the latter part of the 20th Century, has fundamentally changed the world of maps (Plewe, 2007). Currently, a very large number of geographic applications require integration over the Internet (Chang and Park, 2004, Nogueras-Iso et al., 2005, Fils et al., 2009). Specifically, the Web Map Service (WMS) (de la Beaujardière, 2006), Web Feature Service (WFS) (Vretanos, 2005), and Web Coverage Service (WCS) (Whiteside and Evans, 2008) have dominated this Internet delivery of geographic information. WMS is a portrayal service, whereas WFS and WCS are geographic data services. Due to a simple design, small number of API functions, and a clear definition of these functions, WMS is now the de facto standard for corresponding modules of mainstream GIS software packages. More than 1000 WMS instances are deployed and maintained by companies, governmental sectors, organizations, and university institutions. By combining layers of spatially referenced data with remotely sensed aerial or satellite images, these high-tech geographers have turned computer mapping into a powerful decision-making tool (Gewin, 2004).

For example, during the recovery stage of a flood disaster, decision makers need to assess damage information about the flooded area. In addition to real-time monitoring information, such as critical infrastructure and water level, background information is needed for effective decision-making. These background data include the use of a Digital Elevation Model (DEM), Land use maps, Transportation networks, Populated points, weather forecasts, etc. The background information can be directly adopted from WMS resources available on the Internet to share and reduce the costs of acquiring and maintaining these datasets.

The distribution of geospatial data via the Internet is becoming widespread, but there are many barriers to simple and effective access to geospatial data online (Virrantaus et al., 2009). For this flood disaster scenario, the following issues may arise:

• As there are so many land use layers available on the Internet, which one is the most promptly served layer? Which one provides the most useful information for this application?

• Should the application server bind to a specific land use layer, dynamically bind to the best land use layer at run-time, or simply provide a series of potentially available land use layers for the users to select?

• If a land use layer is bound to an application, how can we deal with the situation that this layer is technically unavailable at run-time? Alternatively, how can we guide users to dynamically select the best land use layer?

These questions are not easily answered. The detailed statistical results of our empirical study of WMS resources in Section 2 further reveal the difficulties and significance of dealing with these issues. Given the current situation of WMS resources, these results demonstrate that it is not easy to obtain a stable combination of map layers from multiple servers.

From a technical point of view, the difficulties of composing a stable map of layers from multiple servers can be easily explained. As an implementation of SOA in geospatial interoperability, OGC Web Service (OWS) architecture observes fundamental principles of SOA based on open standards. Fig. 1 is an interactive model demonstrating these roles in OWS architecture.

When this model is applied to a real application, the application is not robust due to the inconsistency between the actual service and the declared service in the registry. This inconsistency may arise due to the following situations:

• The information about a service is defined in a capability document, and the system is not sure that this capability document is exactly what the service provides.

• The metadata of the service provider is in a registry. This registered information may not coincide with the actual service. Alternatively, the service is updated while the metadata is not updated.

• A service may be registered in several registries that support different standards and reference models. Transformation between these different reference models may generate inconsistency.

• A deployed service may cease to work for technical reasons (e.g., power failure, corruption of a supported database, and version conflict between modules), but the registry does not typically reflect these interruptions.

As a result, inconsistency between the metadata in the registry and the actual service is very common. Even when inconsistency does not exist, some servers may perform at low efficiency due to other factors. There are many technologies to improve service efficiency, such as cluster and multithread (Yang et al., 2005), Web Map Tiling Service (WMTS), and Web Map Service-Cached (WMS-C). However, all of these services are under control of the host, and clients can do nothing to enhance performance. Network bandwidth and stability also significantly affect the user experience. All of these factors are research considerations in determining the Quality of Geospatial Information Service (QoGIS, Wu and Zhang, 2007, Wu et al., 2005).

However, rich information resources are accessible from various servers with the increasing number of WMS resources available over the Internet. For example, a search for the keyword “elevation” engages 25 servers and 72 layers including http://wms.jpl.nasa.gov/wms.cgi http://giswebservices.massgis.state.ma.us/geoserver/wms. As a result of this process, selection of more valuable and informative service by clients becomes a prominent issue for map composition. Clients face the issue of choosing or creating a composition plan among several possible plans that will satisfy their quality-of-service (QoS) requirements (Ko et al., 2008). However, clients do have the choice of selecting better service while unable to improve these services themselves.

Herein, this paper proposes a system architecture for integrating geographic information services that supports quality selection. This design includes a continuous operating module to monitor and evaluate the quality of registered geographic information services. The result of this monitoring, evaluation, and the associated historical information are stored in the registry. An interface is provided for a client to query the quality of a service online. Based on this interface, any client can bind the best run-time service for best performance.