1. INTRODUCTION

The US National Science Foundation has initiated a trans-formation of ocean science with the Ocean ObservatoriesInitiative (OOI). The OOI is designed to provide new, per-sistent, interactive capabilities for ocean science, and hasa global physical observatory footprint. The OOI Inte-grated Observatory comprises Regional Scale Nodes(RSN) and Coastal/Global Scale Nodes (CGSN) provid-ing cabled and buoy observatories with mobile instrumentplatforms, respectively.

The OOI Cyberinfrastructure (CI) constitutes the inte-grating element of the OOI Integrated Observatory and isextensible to promot2 platform and sensor expansion inthe 25-30 year future of the OOI. The CI links and bindsthe physical observatory, computation, storage and net-work infrastructure into a coherent system-of-systems.The core capabilities and the principal objectives of theOOI Integrated Observatory are collecting near-real-timedata, analyzing data, modeling the ocean on multiplescales and enabling adaptive and interactive experimen-tation within the ocean. A traditional data-centric CI, inwhich a central data management system ingests data andserves them to users on a query basis, is not sufficient toaccomplish the range of tasks ocean scientists will engagein when the OOI is implemented. Instead, highly distrib-uted collections of capabilities are required that facilitate:

• End-to-end data preservation, provenance mainte-nance and access,

• End-to-end, human-to-machine and machine-to-machine control of how data are collected and ana-lyzed,

• Direct, closed loop interaction of models with the dataacquisition process,

• Virtual collaborations created on demand to drivedata-model coupling and share ocean observatory re-sources (e.g., instruments, networks, computing, stor-age and workflows) – the Virtual Observatory,

• End-to-end preservation of the ocean observatoryprocess and its outcomes, and

• Automation of the planning and prosecution of ob-servational programs.

The OOI CI provides the software services, the user in-terfaces to support these applications; in addition it pro-vides the underlying integration infrastructure comprisingmessage-based communication, governance and securityframeworks, similar to the role of the operating system ona computer. Users or system managers are able to com-pile new, specialized observatories through the CI.

2. SCIENTIFIC PROCESS VIEW OF THE OOI CI

From the outset, the OOI CI was designed to implementa set of activities derived from a process model that struc-tures their interaction and compartmentalization (Fig. 1).The model supports the real-time coupling and aggrega-tion of the main observe, model and exploit activities aswell as their respective sub-activities.

The observe activity constitutes the data collection strat-egy that lies at the core of scientific investigation. Theoutcome is a set of observational products such as data orquantities derived from them that serve as the input to amodel activity. The result of modeling is a set of derivedproducts that yield an interpretation of the data and theprocesses that determine them. Based on this improved

CYBERINFRASTRUCTURE FOR THE U.S. NSF OCEAN OBSERVATORIES INITIATIVE: A MODERN VIRTUAL OBSERVATORY

Orcutt, J.(1,2), Vernon, F.(1), Peach, C.(1), Arrott, M.(2), Farcas, C.(2), Farcas, E.(2), Krueger, I.(2), Meisinger, M.(2), Chave, A.(3), Schofield, O.(4), Kleinert, J.(5)

(1) Scripps Institution of Oceanography, University of California, San Diego, La Jolla, CA, 92093-0225, USA(2) California Institute of Telecommunications & Information Technology, University of California, San Diego,

La Jolla, CA, 92093-0436, USA(3) Woods Hole Oceanographic Institution, Woods Hole, MA, 02543, USA

(4) Rutgers University, Princeton, NJ, 08901, USA(5) Raytheon Intelligence and Information Systems, Aurora, CO, 80011, USA

Figure 1. CI Process Model

understanding of the underlying physics, chemistry andbiology, the exploit activity is used to plan, simulate andexecute additional observation activities. The entire sci-entific process operates as a closed rather than an openloop, with or without human intervention. The integrateddata, model output, data analysis and resource exploita-tion planning/execution all become elements of a knowl-edgebase, a great step beyond the more familiar database.

3. FUNDAMENTAL STRATEGIES

The CI integration strategy is based on two core princi-ples: messaging and service-orientation. A high-perfor-mance message exchange provides a communicationconduit with dynamic routing and interception capabilitiesfor all interacting elements of the system-of-systems. Themessage interface is isolated from any implementationtechnologies, and provides scalability, reliability andfault-tolerance. Service-orientation is the key to managingand maintaining applications in a heterogeneous distrib-uted system. All functional capabilities and resources (e.g.sensors, storage, gliders, buoys or cables) represent them-selves as services to the observatory network, with pre-cisely defined service access protocols based on messageexchange. Services are also defined independently of im-plementation technologies. Assembling and integratingproven technologies and tools using the integration infra-structure provide the functional capabilities of the CI.

The CI deployment strategy is based on the concept of acapability container (Fig. 2) that provides all essential in-frastructure elements and selected deployment-specificapplication support. Capability containers can be de-ployed wherever CI-integrated resources are required

across the observatory network, and can be adapted toavailable resources and their environment. This includes(for example) platform controllers on remote, intermit-tently connected global moorings, compute units placedin the payload bay of AUVs and gliders and the full rangeof terrestrial CI deployments (Cyber Points of Presenceor CyberPoPs). The CI multi-facility strategy supportsthe collaboration of multiple independent domains of au-thority bringing together their own resources, with nocentral governance and policy authority, based on agree-ments and contracts. This enables the sharing and use ofobservatory resources across the network, governed byconsistent policy.

4. SERVICES NETWORKS AND SUBSYSTEMS

The CI’s functional capabilities are structured into sixservices networks (SNs) that participate in operational ac-tivities, resulting in services that support user applications.The application-supporting services networks are Sens-ing and Acquisition, Data Management, Analysis andSynthesis, and Planning and Prosecution. The infrastruc-ture services networks are Data Management, CommonExecution Infrastructure (CEI), and Common OperatingInfrastructure (COI). Data Management provides both ap-plication and infrastructure services.

The services networks (SN) and their constituent servicesare implemented and integrated into subsystems of thesame name. Subsystems are the primary units of workbreakdown and will be delivered by individual subsystemconstruction projects.

The Sensing and Acquisition SN performs instrumentmanagement, mission execution, and data acquisitiontasks. The Instrument device model comprises one ormore physical sensors or actuators, and is represented inthe CI by a logical device, the Instrument Agent. Thephysical device provides sensor data and status informa-tion to the Instrument Agent, and receives configurationinformation and commands from it. Each InstrumentAgent has an Instrument Supervisor that is responsible formonitoring the state of health of the Instrument. Each plat-form has a Platform Agent that is responsible for control-ling all devices on that platform. ObservatoryManagement is responsible for coordinating all observa-tory resources to ensure their safe operation. The Ex-change service is a projection of COI capabilities, withthe main purpose of establishing a publish-subscribemodel of communication that ensures distributed data de-livery to all end-points. It appears in all SN’s.

The Instrument Agent is responsible for translating com-mands from the Platform Agent to the Instrument, eventhandling, acquiring data from the Instrument, managingthe state of the Instrument, providing direct access (e.g.,ssh) to the Instrument, and updating the Instrument clockFigure 2. Capability Container

from an external source. The Instrument Supervisor re-ceives status information

5. THE OOI AS A SYSTEM OF SYSTEMS.

The primary goal of the Integrated Observatory Cyberin-frastructure functional design is to support the activitiesand applications of:

• Scientific Investigation, supporting researchers in thestudy of environmental processes though observa-tions, simulation models and expressive analyses andvisualizations, with results that directly feed back toimprove future observations.

• Education and Participation, supporting education ap-plication developers, educators and the general publicfor accessing and understanding OOI resources inways suitable for specific target audiences.

• Community Collaboration, enabling OOI users toshare knowledge and resources, and to work togetherin project settings and ad hoc communities.

In support of these activities, a variety of Integrated Ob-servatory resources of different type and purpose need tobe administered, including:

• Observation Plans, providing activity sequences, serv-ice agreements and resource allocations for observa-tional campaigns, and similar templates forevent-response behaviors;

• Data Sets, representing observational and derived dataand data products in the form of data archives andreal-time continuous data streams;

• Processes, representing data collection and process-ing workflows that arrange multiple steps involvingmultiple actors and resources;

• Instruments and marine observatory infrastructure el-ements, such as telemetry systems, GPS and data log-gers;

• Models, including numerical ocean forecast modelsand their configurations, as well as other analysis andevent detection processes;

• Knowledge, representing all metadata, ancillary data,analysis results, association and correspondence linksbetween resources, and knowledge captured in on-tologies for semantic mediation purposes.

The support for these activities and resources rests on acollection of infrastructure services that provide resourcemanagement, interaction, communication and process ex-ecution. The CI Capability Container is the extensible,deployable base unit of CI capabilities. The IntegratedObservatory’s functional capabilities are structured intosix services networks (i.e., subsystems) that were dis-cussed earlier: four elements that address the ocean andEarth science- and education-driven operations of the OOIintegrated observatory, and two elements that provide coreinfrastructure services for the distributed, message-based,

service-oriented integration and communication infra-structure and the virtualization of computational and stor-age resources.

The Sensing and Acquisition services network providescapabilities to interface with and manage distributedseafloor instrument resources, as well as provide qualitycontrol services. The Data Management services networkprovides capabilities to distribute and archive data, in-cluding cataloging, versioning, metadata management,and attribution and association services. The Analysis andSynthesis services network provides a wide range of serv-ices to users, including control and archival of models,data analysis and visualization, event detection servicesand collaboration capabilities to enable the creation of vir-tual laboratories and classrooms. The Planning and Pros-ecution services network provides the ability to plan,simulate and execute observation missions using taskableinstruments; it is the CI component that turns the OOI intoan interactive observatory.

The remaining two services networks are the CommonExecution Infrastructure (CEI) and the Common Operat-ing Infrastructure (COI). The CEI provides an elasticcomputing framework to initiate, manage and storeprocesses that may range from initial operations on dataat a shore station to the execution of a complex numericalmodel on the national computing infrastructure and oncompute clouds. The COI includes capabilities to manageidentity and policy, manage any resource’s life cycle, aswell as catalog and repository services for observatory re-sources. An efficient messaging and service bus that in-corporates security and governance, and providesguaranteed delivery, lies at its heart. Service-orientationand messaging realize loose coupling of components, re-sulting in the flexibility and scalability that are key in sucha complex large-scale system-of-systems. All OOI func-tional capabilities and resources represent themselves asservices to the observatory network, with precisely de-fined service access protocols based on message ex-change. Fig.2 depicts a capability container, indicated bythe octagon shape, with interfaces to local resources andto the network environment. Local resources includephysical resources such as instruments (sensors) and ma-rine observatory infrastructure, storage resources such asdisks and network drives, and computing resources suchas grid nodes, cloud computing instances, and CPUs onmobile platforms such as AUVs (Autonomous Underwa-ter Vehicles). Capability container can also be connectedto user interfaces, external applications and to capabilitycontainers in different, independent facilities that havetheir own domains of authority and operation. No matterwhere deployed, the capability container provides all ofthe infrastructure and application support required for aninstallation site within the OOI Integrated Observatorynetwork. The capability container hosts the six services

networks and their resource interfaces as depicted in thefigure. The footprint of a capability container can vary de-pending on the resource constraints of its hosting envi-ronment. The selection of functional capabilities presentin a specific capability container depends on the respec-tive needs and resource availability at this specific loca-tion in the network. For instance, on an intermittentlyconnected instrument platform, instrument access, dataacquisition and data buffering capabilities provided by theSensing and Acquisition services are required, while theAnalysis and Synthesis capabilities are not. In contrast, atthe core installation sites, data processing, numericalmodel integration and event response behaviors need to bepresent.

6. SUMMARY

The Ocean Observatories Initiative faces the enormouschallenge of building a cohesive distributed system-of-systems that incorporates a large number of autonomousand heterogeneous systems, deals with instruments andcomputational resources of a wide range of capabilities,serves the needs of diverse stakeholders, and accommo-dates change over the timescale of decades. A carefullythought out architecture is key to addressing this chal-lenge. We find that simplicity wins and a few core princi-ples help us organize the OOI properly.