RedCLARA built a continental backbone through federation, not ownership

Summary

• RedCLARA is a Uruguayan non‑profit association that federates national research and education networks; it is neither a shareholder company nor an operator that controls every campus.
• Eight main routed nodes and long‑term international capacity form a 100G‑class backbone, with declared links of 100, 200 and 300 Gbps and a smaller Central American segment.
• ALICE and ALICE2 created the institution and the first network; BELLA secured lasting capacity on the direct Portugal–Brazil route, reducing dependence on indirect paths via North America.
• The test of BELLA II is converting project funding into lasting, member‑sustained infrastructure, reinforcing weak national networks, public evidence and supplier diversity.

BELLA reshaped the network’s geography

BELLA, Building the Europe Link to Latin America, combined a submarine segment with terrestrial upgrades. BELLA‑S secured long‑term optical spectrum on a direct cable between Portugal and Brazil. BELLA‑T reinforced the South American path needed to carry that capacity from the landing to regional nodes and national networks.

The physical system became EllaLink, a roughly 6,000‑kilometre route between Sines and Fortaleza. Research and education traffic entered production in 2021. Two 100 Gbit/s links were initially enabled: one between GÉANT and RedCLARA and another for Copernicus. The 2024 report describes a broader BELLA infrastructure of 35,830 kilometres and Nx100G capacity, covering much more than the submarine segment.

Lengths and costs measure different perimeters. CORDIS records €12.4447 million for the BELLA‑S1 action, including €10 million from Europe, while other totals include terrestrial works, additional financing lines and national contributions. They must not be merged into a single figure without explanation.

Programme sources have mentioned latency reductions of up to 50 % for certain paths. The claim must remain tied to specific routes; it is not a universal result. The lasting change is the existence of a direct South Atlantic route and expandable rights as optics and demand evolve.

Forty optical channels do not equal owning the cable

BELLA obtained long‑term rights over forty optical channels on EllaLink. A spectrum right allows control over how capacity is lit and upgraded, but it does not make RedCLARA or GÉANT the exclusive owners of the commercial cable, the landing stations or all ancillary facilities.

The chain runs from the fibre pair to the spectrum right, from that to the optical channels, from the channels to 100G or later coherent transmission, and from the station to terrestrial networks. The cable operator maintains the marine system; landing and colocation facilities have other responsible parties; RedCLARA and GÉANT coordinate the academic capacity; manufacturers supply equipment; NRENs deliver the traffic to institutions.

Long‑term spectrum can be more valuable than a fixed circuit because throughput can grow by swapping equipment within optical and contractual limits. At the same time it exposes the consortium to shared‑cable governance, repairs, power, spectrum planning and interoperability. The right is durable, not self‑sufficient.

This ownership boundary is essential. RedCLARA can legitimately speak of a regional backbone and a transatlantic route because it coordinates and operates capacity. It should not be described as owning every line on the map. Its power comes from federation and contractual control of strategic capacity, not from complete physical integration.

BELLA and the direct South Atlantic route

Before BELLA, much of the research traffic between Europe and Latin America followed paths via North America. The detour reflected the economics of cables and interconnection, but it added distance, concentrated risk and left the scientific community dependent on routes bought for other markets.

BELLA tackled the problem in two parts. BELLA‑S acquired spectrum rights on a direct submarine connection. BELLA‑T reinforced the South American terrestrial network so that capacity reaching Fortaleza could travel to regional nodes, NRENs and institutions. A cable making landfall does not by itself create a continental network; traffic must travel hundreds or thousands of additional kilometres.

EllaLink connects Sines and Fortaleza over approximately 6,000 kilometres. Production academic traffic started in 2021. Two initial 100 Gbit/s links were enabled: one GÉANT–RedCLARA and another for Copernicus. The direct route provides an alternative to North America and can improve latency and throughput for specific pairs.

Sources have mentioned reductions of up to 50 % on certain paths. It must not be generalised: the outcome depends on origin, destination, national access, routes and the remote network. The structural change is clear: there is a direct route, scalable capacity and greater geographic diversity.

North American paths remain necessary. Miami continues as an international gateway and can serve as a detour when the South Atlantic fails. Resilience comes from combining routes, not from replacing one dependency with another. BELLA is both a performance improvement and a diversification decision.

Spectrum rights, terrestrial reach and the ownership boundary

BELLA’s forty optical channels give RedCLARA and GÉANT lasting rights over part of the EllaLink spectrum. They can be lit and upgraded with new generations of coherent transmission. This model is more flexible than fixed capacity, because terminal equipment can increase throughput within optical and contractual limits.

The right does not equal full ownership. EllaLink remains a shared commercial system. Stations, buildings, power, marine repair, spectrum planning, equipment and terrestrial tails fall under different responsibilities. RedCLARA coordinates and operates strategic capacity without owning the entire physical chain.

BELLA‑T demonstrates why the terrestrial leg is as important as the submarine one. From Fortaleza, traffic must reach São Paulo, Porto Alegre, Buenos Aires, Santiago, Manta and other points. Fibres may be owned by NRENs, leased, contributed in kind, or protected by distinct agreements. The map combines those rights without becoming a title register.

The architecture delivers collective power, but it multiplies dependencies. A cut, a failed station, a contractual dispute or obsolete equipment can reduce the spectrum’s value. Long‑term contracts must cover maintenance, upgrades, continuity and exit, not just nominal throughput.

Financial numbers must respect the same boundaries. The cost of BELLA‑S1, European contributions, terrestrial infrastructure and in‑kind valuations are not RedCLARA’s annual revenue. A large figure can represent assets spread over years, not cash available to the association.

A continental network that is not a single network

RedCLARA invites confusion when it is described simply as “the Latin American research network.” The phrase suggests a single operator with direct control from the international backbone to a university port. The real architecture is federated. A researcher usually reaches RedCLARA via an institutional network and then a national research and education network, or NREN. RedCLARA provides the regional and intercontinental layer that allows those separately governed systems to work within a wider scientific environment.

The distinction defines both the organisation’s value and its limits. RNP in Brazil, REUNA in Chile, RENATA in Colombia, CEDIA in Ecuador, CUDI in Mexico, RedCONARE in Costa Rica and RAU in Uruguay each keep their national governance, domestic infrastructure, relationship with institutions and operational responsibilities. RedCLARA does not replace them. It interconnects them, represents common requirements, coordinates capacity and services, and opens access to peer networks such as GÉANT, Internet2 and CANARIE.

An end‑to‑end path can cross a campus LAN, an institutional firewall, a national circuit, the NREN backbone, RedCLARA, an international peer, the destination NREN and a remote facility. A 100 Gbit/s regional leg does not compensate for a congested campus edge, slow national access, misconfigured storage or restrictive policies at the destination. Conversely, a single national network cannot negotiate every intercontinental route, identity relationship or collaborative service that modern science demands.

The useful description is a shared layer of coordination and infrastructure. “Shared” refers to governance, engineering, long‑term capacity rights and common services, not ownership of every fibre, router, point of presence or cable segment drawn on a regional map. It is not a trivial terminological point: it is the central mechanism by which RedCLARA turns multiple national systems and commercial contracts into a regional network of public interest.

Why Latin America needed a regional layer

The foundational problem was fragmentation. In the early 2000s, several countries had national or emerging academic networks, but the region lacked a lasting continental organisation capable of aggregating demand, operating a regional backbone or negotiating on equal footing with European and North American networks. International circuits were expensive, national telecom markets were very different, and commercial routes followed the economics of operators, not the needs of distributed science.

Geography imposed practical penalties. A university could have good domestic connectivity while depending on expensive, indirect international transit to collaborate with another Latin American institution. Traffic between Europe and Latin America could travel via North America because the commercial structure of cables and interconnection made buying that path simpler. The outcome was not only latency: it also meant weaker bargaining power, limited route diversity and dependence on services whose priorities were set outside the scientific community.

A centralised continental operator would not have been realistic. The region showed large differences in NREN maturity, public funding, regulation, fibre availability, university systems and landing points. National organisations needed to retain legal and operational autonomy. The viable model was a federation: domestic networks would keep their national role, and a regional association would coordinate the common layer and global relationships.

The design also created political and institutional value. An isolated NREN can struggle to influence an international programme or justify a specialised route. A regional organisation can pool scientific use cases, combine contributions, link grants with national assets and present itself as a coherent interlocutor to the European Commission, GÉANT, development banks, operators and cable companies. RedCLARA was born to turn that collective need into an institution and an operational network.

Toledo, Valle de Bravo and the legal association

The institutional idea took shape in 2002. Representatives of Latin American NRENs debated the feasibility of a regional network during a CAESAR meeting in Toledo, Spain, and continued the work in Rio de Janeiro, Buenos Aires, Santiago and Mexico. They had to solve not only topology, but who would participate, how the regional level would relate to national networks, how decisions would be made, and how an externally funded project could turn into a lasting organisation.

The legal chronology contains three dates that must be kept. RedCLARA’s history states that the statutes were signed in Valle de Bravo, Mexico, on 9 June 2003. A civil society register of the Organization of American States uses 10 June 2003 as the establishment date. RedCLARA further states that its legal existence was recognised under Uruguayan law on 23 December 2003. The evidence allows them to be treated as related but distinct events: the constitutive agreement, administrative registration and later legal recognition in Uruguay.

The best‑supported description is that of an international non‑profit civil association, recognised under Uruguayan law and headquartered in Montevideo. RedCLARA sometimes uses the phrase “international law organisation” in English, but the available statutes and registers do not prove it was created by treaty or that it is an intergovernmental organisation. It is a membership association with regional aims, not a supranational regulator or a public agency.

The current statutes were approved in November 2019 and published in January 2020. They establish the General Assembly as the supreme body, a Board of Directors, a Fiscal Commission, a Technical Commission that gathers NREN expertise, and an Executive Directorate responsible for implementation. Legitimacy comes from association and cooperation, not from shareholder ownership or territorial authority.

Seven members, eleven connected countries and a wider mission

The current public directory lists seven NREN members: RNP, RENATA, RedCONARE, REUNA, CEDIA, CUDI and RAU. The advanced connectivity page, however, states that the network connects eleven countries: Argentina, Brazil, Chile, Colombia, Costa Rica, Ecuador, El Salvador, Guatemala, Mexico, Paraguay and Uruguay. Older material mentions thirteen. These figures are not interchangeable.

Legal membership determines governance rights, obligations and associational participation. Operational connectivity can exist through partners, temporary arrangements or projects even if the national NREN is not a voting member. Historical totals may reflect former memberships or old programme footprints. The accurate account distinguishes seven current legal members from eleven connected countries and treats older numbers as dated data.

The difference creates a strategic tension. RedCLARA aims to serve a wider region than the association that formally governs it. A country can benefit from connectivity or from BELLA II without holding an equivalent seat in the Assembly. Conversely, the seven members bear a disproportionate share of governance, technical work and perhaps recurring funding. It does not prove failure, but it affects legitimacy, cost‑sharing and the building of a genuinely regional platform.

The 2025‑2027 Board reflects active membership. José Palacios, from REUNA, is president; Eduardo Grizendi, from RNP, vice‑president; Juan Pablo Carvallo, from CEDIA, treasurer; Carlos Gamboa, from RedCONARE, secretary; and Eduardo Grampín, from RAU, member. Luis Eliécer Cadenas is executive director. NRENs keep formal control, but public information does not provide recent Assembly minutes, voting weights, current dues or a full explanation of how non‑member connected countries participate.

ALICE provided the start‑up capital for a regional network

The legal formation of RedCLARA and its first operational network were inseparable from ALICE, the Europe–Latin America interconnection project. The European Commission and DANTE signed the contract in June 2003. The programme had €12.5 million: €10 million from the Commission and €2.5 million from Latin American partners. It ran until March 2008 after several extensions.

ALICE solved several start‑up problems at once. It provided capital, created common procurement, gave the new association an immediate operational mission and connected the region with GÉANT. Instead of waiting for a membership‑fee revenue model to mature, it allowed the network’s value to be demonstrated while the institution was being consolidated.

The first infrastructure was modest compared with today’s, but its importance was institutional. It established regional points and production connections between NRENs. It also linked Latin America with European and North American networks, creating technical relationships that would later extend to identity, middleware, grid computing, open science and shared projects. The working network turned the association into more than a forum.

ALICE built in a difficult question: what happens when a temporary grant finances infrastructure that must operate indefinitely? Contracts, equipment, staff and expectations do not disappear when a project report closes. RedCLARA’s subsequent history can be read as a search for longer asset horizons—IRUs, fibre, wavelengths, spectrum, services and co‑investment—that keep the institution alive between cycles.

ALICE2 moved coordination to Latin America

ALICE2 began in December 2008 under an €18 million European Commission contract. The decisive institutional change was that RedCLARA took over coordination of the programme, rather than participating under European coordination. It was a transfer of operational and project leadership to the Latin American association.

The project explicitly sought sustainability: lower recurrent costs through dark fibre, wavelengths and IRUs; wider coverage; stronger NRENs; technical training; user communities; and more stable funding. Sources disagree between an end in 2012 and an administrative closure in January 2013. The safest interpretation is that operational activities concluded before the administrative closure.

ALICE2 broadened the definition of infrastructure. Grid initiatives connected regional resources with international collaborations. Identity and middleware projects facilitated authentication and access to common services. Training recognised that a fibre route has little value if national networks lack personnel, governance or applications capable of using it.

The programme took RedCLARA from a young transport infrastructure towards a regional e‑infrastructure organisation. It still depended on external funding, but it gained experience in coordinating large projects, managing technical work across countries and turning transport into a platform for other services.

From recurrent circuits to long‑term capacity rights

The economics of a regional network depend on how capacity is acquired. A short‑term managed circuit offers a defined service, but it exposes the operator to price renegotiations and limits control over upgrades. A lasting right over fibre, a wavelength or spectrum demands greater complexity, but it allows longer‑term planning and the ability to upgrade transmission without re‑purchasing every bandwidth increment.

RedCLARA’s strategy increasingly favoured IRUs, dark fibre, wavelengths, optical channels and capacity swaps where possible. The organisation did not become a conventional owner of all cables, but it acquired rights capable of surviving a technological generation. Its 2015 report linked this approach with lower variable costs and greater sustainability.

Long rights also change the risk. They reduce dependence on recurrent purchases, but they increase exposure to the underlying cable, landing stations, operator, maintenance and contract. Spectrum is only worth something if the system is repaired, powered and operated. Optical equipment, colocation, terrestrial tails and skilled staff are still needed.

That is why the asset base does not fit a simple ownership map. Some fibre may belong to an NREN, some may be leased, contributed in kind, covered by an IRU or represented by optical rights in a shared cable. The regional platform is assembled from legal and technical relationships as much as from physical assets.

Before BELLA, Europe often meant North America first

The absence of a direct high‑capacity route between Europe and Latin America defined both performance and dependency. Even between institutions on opposite sides of the South Atlantic, traffic could follow commercial systems via North America. The path made sense for operators, but it added distance, concentrated transit and limited the scientific community’s control over capacity.

North American interconnection was, and remains, essential. Initiatives such as WHREN/LILA strengthened routes between Latin American networks and the United States, including links via Tijuana, San Diego and Pacific Wave. Miami became a gateway to Internet2, CANARIE and other partners. Those paths provide reach and are part of resilience.

The problem was dependence, not the existence of the routes. A continental network needs alternatives. A direct connection to Europe could reduce latency for certain pairs, diversify failure modes and support large scientific flows without forcing all traffic through the same corridor. It also aligned physical geography with the long‑standing cooperation between RedCLARA and GÉANT.

The need grew with Earth observation imagery, genomics, telescope data, climate models and HPC workloads. BELLA turned the aspiration for a direct route into a long‑term physical and contractual system.

The current topology: eight nodes and several capacity classes

RedCLARA’s current technical description identifies eight main routed nodes: Buenos Aires, Fortaleza, Porto Alegre, São Paulo, Santiago, Manta, Panama City and Miami. The map groups different functions. South American networks enter a regional ring; Panama City aggregates and extends Central American coverage; Miami is a main hand‑off to Internet2, CANARIE and other North American relationships; and Fortaleza is the Atlantic node for the direct route to GÉANT and other research partners.

The South American ring is described as 100 Gbit/s or greater. Another current page speaks of 100, 200 and 300 Gbit/s links. The shared Central American infrastructure is placed at 20 Gbit/s. These are not incompatible answers: a backbone class, an aggregated or protected link, a regional subnet and an optical channel can present different capacities at the same time.

The safest formulation is that RedCLARA operates a 100G‑class regional network, with some links described at 200 or 300 Gbit/s, while Central American capacity is lower. An old mention of 10 Gbit/s remains on the site, but it contradicts the current topology and seems to be documentary debt. It should not be used to downplay the current network, just as the largest published figure should not be applied to every path.

Capacity also changes by layer. An optical system can support several 100G wavelengths. An interface can be protected by two links. A service can be limited below the port speed. An NREN may receive less capacity than the core. University access can be the bottleneck even when the international route has headroom. Figures are only useful when their scope is explained.

The same caution applies to distance. The Portugal–Brazil submarine segment is approximately 6,000 kilometres, while the 2024 report gives 35,830 kilometres for the expanded BELLA infrastructure. The latter includes terrestrial and international components. Both can be correct, but neither should be published without defining the measured entity.

A logical diagram also does not prove physical diversity. Two paths can share ducts, landing stations, carrier buildings or power. The ring shape suggests alternative routing, but resilience depends on contracts and civil works that the map does not show. RedCLARA coordinates some of those hidden layers, although it does not publish a full asset and per‑route diversity inventory.

AS27750, peering and separation of traffic classes

AS27750 is RedCLARA’s public autonomous system identity. Through it, the organisation carries or exchanges routes from member networks, global research peers, service infrastructure and selected commercial connectivity. The layer supports IPv4, IPv6, multicast, direct peering and protected regional paths; it is how the federation appears to the global routing system.

Public databases show relationships with GÉANT, Internet2, CANARIE, ESnet, RNP, REUNA, RENATA, RedCONARE, CEDIA and TENET. PeeringDB describes a selective policy and presence in important environments. These records are operational snapshots, not contractual maps. They can demonstrate route exchange without revealing economic terms, committed capacity, protection or who supplies the physical circuit.

Peering can reduce distance, latency and transit cost, but it demands ongoing work. Prefix filters must be accurate; maximum‑prefix limits must accept legitimate changes without letting a leak overwhelm the session; IRR and RPKI information must be maintained; asymmetric paths complicate diagnosis. A direct BGP session is an engineering relationship, not a cable that is connected once.

RedCLARA also separates academic traffic, commercial traffic and private Layer 2 circuits. The separation allows eligible destinations to use specialised routes while general Internet traffic follows other peers or transits. NRENs can aggregate commercial demand without losing research‑service characteristics.

The separation is logical and operational, not physical isolation. The same router, fibre, rack, cable or building can support several classes. RedCLARA is interconnected with the commercial Internet and offers commercial services; it does not exist in a sealed universe. Its differentiation comes from eligibility, routing policy, dedicated capacity, community trust and service design.

This precision also affects claims of sovereignty. An academic route can reduce dependence on the public Internet for certain workloads, but it does not eliminate fibre owners, colocation, manufacturers or cable companies. The mission changes who coordinates the path and for what purpose; it does not erase the physical supply chain.

Layer 2 circuits, operations and the boundary of regional responsibility

RedCLARA offers dedicated Layer 2 virtual circuits. A VLAN can make laboratories, data centres or instruments in different countries work as if they share an Ethernet environment. It can be used to move data, extend a private cloud, link an observatory with remote computing or support a test that should not depend on ordinary public routing.

Layer 2 extension also expands the failure domain. Broadcast storms, spanning tree, unknown unicast flooding, MTU mismatch, weak segmentation or identifier errors can cross multiple networks. RedCLARA supplies the regional segment, but NRENs and institutions must coordinate their parts, their equipment, their controls and the responsibility boundaries.

The same holds for performance. An institution can buy a fast regional circuit and still be limited by its local equipment, a firewall or a national link. The incident may lie in the campus, the NREN, RedCLARA, the international peer or the destination. A shared service does not eliminate the plurality of administrative domains.

RedCLARA organises this layer through three groups. NEG plans architecture and capacity; the NOC supervises the network and coordinates incidents; SEG implements and operates value‑added services. The division links long‑term planning, daily operation and software platforms.

Public information is stronger on structure than on outcomes. There is no complete incident database, mean‑time‑to‑repair, route leaks or per‑service SLA methodology. The old claim of 100 % average monthly availability does not define measurement points, maintenance or scope. A mature federation should publish telemetry and common definitions so that failures can be located and end‑to‑end performance assessed.

The responsibility boundary is not an accidental defect; it is part of the federation. RedCLARA must define interfaces, maintain escalation channels and coordinate actors without pretending it controls every piece of equipment. Quality depends on handoffs, contacts and shared telemetry as much as on core throughput.

BELLA II: expansion through co‑investment, not a finished map

BELLA II was agreed in December 2022 for forty‑eight months. The European Commission committed €13 million within a €28 million plan, and RedCLARA must mobilise at least €15 million from governments, development institutions, firms, banks and other partners. The organisation executes and coordinates the programme.

Objectives include connecting or reinforcing Peru, Costa Rica, El Salvador, Guatemala and Honduras, with possible extensions to Mexico, the Caribbean, Paraguay, Uruguay, Bolivia and Belize. The 2025 competitive dialogue sought a system of at least fifteen years, with initial minimums of 20 Gbit/s in Central America and 100 Gbit/s in Peru, designed for growth.

These are objectives and requirements, not proof that all routes were active on 3 August 2026. Memoranda, dialogues and announcements show institutional preparation; they do not demonstrate technical acceptance, production traffic or closed funding. The account must keep the future tense where evidence is prospective.

The competitive dialogue recognises that a one‑size solution does not fit the whole region. One country may need an IRU, another a swap, another joint construction or a managed service. Suppliers and investors help define architecture, financing, maintenance and operation. Flexibility can improve the outcome, but it complicates cost comparison, timeline and responsibilities.

BELLA II will be measured by commissioned routes, protected capacity, functional NRENs, sustainable agreements and institutions able to use the service. The expansion is as much institutional development as physical construction.

From a transport network to a regional research platform

RedCLARA’s first generation showed that national research networks could exchange traffic over a regional backbone. Afterwards, the organisation tried to answer a harder question: what shared capabilities become possible once that network exists? Its portfolio today includes identity, file transfer, collaboration environments, training, high‑performance computing access, open science, cybersecurity coordination and thematic communities.

This shift does not abandon networking; it multiplies its value. A high‑capacity route has limited impact if researchers cannot authenticate to a service, locate computing resources, move a large dataset, find partners or get support for a cross‑border workflow. The application and community layers turn raw capacity into usable scientific infrastructure.

BELLA II formalises that platform logic through innovation hubs, testbeds, open calls, training, strategic dialogues and mechanisms for linking projects with partners or funders. The 2024 agri‑food Ideathon, with 125 experts from 28 countries, showed how RedCLARA could organise a regional problem‑solving process. The early HPC testbeds opened a path from training and prototyping toward shared computational resources.

A testbed is not a production service. It usually has a selected cohort, limited capacity, expert support and an experimental goal. It can reveal technical and organisational obstacles before committing permanent infrastructure. It can also create misleading expectations if a successful demonstration is presented as universal availability or commercial maturity.

RedCLARA’s opportunity is to become a federation of services and testbeds, not just a transport network. That would make the regional layer relevant to researchers who never think about BGP or optical channels. It would also increase operational complexity: identity, storage, computing, notebooks, collaboration tools and sensitive data introduce security, privacy, support and lifecycle obligations.

The organisation thus moves up the stack without owning every component. National networks, SCALAC members, universities, cloud providers and project partners still operate the underlying resources. RedCLARA typically integrates, curates, coordinates and provides a regional access route. The platform is federated for the same reason as the network: capabilities are distributed, and no single institution can replicate them all efficiently.

BIO+IA and the discipline of early evidence

BIO+IA Early Adopters is one of the clearest examples of RedCLARA’s new platform role. Opened in 2026 under BELLA II, it selected five projects from six countries to offer training, expert support and access to an artificial‑intelligence and bioinformatics environment. Projects addressed breast‑cancer biology, molecular screening, antimicrobial or mercury resistance, drug repurposing for liver disease, and translational biomedicine.

At the July 2026 Demo Day, participating teams explained how the environment helped them search, integrate and interpret biological information. It is significant evidence of a facilitated research flow. It shows that a regional programme can combine network access, specialist tools and expert guidance to help distributed teams work on data‑intensive biomedical problems.

The evidence does not establish clinical efficacy. A model that identifies a candidate molecule has not proved a treatment works. A bioinformatics pipeline that accelerates analysis has not completed laboratory validation, regulatory review or health‑system deployment. The programme served a limited cohort and its results come from entities and organisers, not from an independent clinical trial.

These cautions do not diminish the programme; they identify the correct unit of achievement. RedCLARA demonstrated that its infrastructure can support a structured regional experiment at the intersection of AI, biology and advanced networking. It lowered barriers to tools, data and expertise for the participating teams, built relationships among researchers from several countries and generated evidence that certain workflows could run in a shared environment.

The next test is continuity. A demonstration turns into infrastructure when teams can return, new cohorts enter, data governance is clear, compute and storage have funding, and successful workflows can be transferred to national or institutional systems. Re‑use, scientific outputs, independent evaluation and post‑pilot funding will show whether BIO+IA becomes a recurring service or remains a valuable but temporary programme.

The general lesson is that RedCLARA must measure innovation by the distance between participation and lasting capacity. Ideathons count people; testbeds count experiments. Infrastructure must also count maintained environments, returning users, governed data, trained operators and projects that continue when the grant or event ends.

Identity as infrastructure: FIEL, eduGAIN and eduroam

A route can work while a researcher cannot log into a repository, supercomputer or laboratory because the service does not recognise their identity. Authentication, attributes and trust metadata are part of access infrastructure.

FIEL supports national identity federations. eduGAIN connects them to an international interfederation. Normally the user goes back to their home institution to authenticate; a signed assertion and selected attributes then reach the remote service. RedCLARA coordinates the federated layer and does not create a continental database of all passwords.

eduroam applies the principle to Wi‑Fi roaming. The visited institution provides local access and the home institution validates the credential over a RADIUS hierarchy. Security depends on the device, home authentication, metadata and the integrity of the national and regional layers.

Federation distributes power and risk. It lets each institution keep identity authority, but a compromised provider, a stolen key or a poor attribute can affect services in other countries. Assurance levels, privacy and response vary. RedCLARA enables cross‑border trust and helps federations mature.

Identity is as critical as routing. A supercomputer that is unreachable because of wrong attributes is as unavailable to the user as if the route were cut. Success depends on the reliability of the trust chain, not just the number of registered institutions.

MiLab, eNVIO and the collaborative research layer

MiLab clearly shows RedCLARA above the network. It combines a Dataverse‑based repository, GitLab for code and projects, Mattermost for communication and a JupyterHub‑type environment with Python and R. Federated identity can bind those components for distributed teams.

The architecture covers a practical cycle: organise the team, communicate, develop software, run notebooks, curate data and publish datasets. For institutions without an equivalent local platform, it lowers the barrier to reproducible collaboration and shows that regional value can also reside in shared software.

MiLab is not a universal production cloud or a substitute for HPC centres. A notebook environment does not offer unlimited computing, certification for any sensitive data, permanent legal preservation or every tool. Storage, availability, privacy, administration and maintenance need continuous governance and funding.

eNVIO solves a narrower problem: moving large files over academic infrastructure rather than consumer services. It can serve data too large for email or ad‑hoc tools, but it is not a continuous instrument pipeline or a long‑term archive.

These platforms add data responsibility. Personal, biomedical, commercial or embargoed datasets need access control, encryption, retention, cross‑border rules and incident response. Federated login is not enough. As RedCLARA operates more platforms, the governance framework will be as important as the software stack.

Ventanilla Abierta and navigating a fragmented ecosystem

The regional e‑infrastructure is distributed across NRENs, computing centres, identity federations, repositories, global partners and project services. A researcher may know the scientific need without knowing which organisation controls storage, HPC allocation, the international route or the required expertise. Ventanilla Abierta works as an orientation layer.

It can direct the request towards connectivity, SCALAC, data, identity, a partner resource or a funding pathway. Its strength is discoverability: it turns independent resources into a more navigable system, without claiming that RedCLARA owns or guarantees everything it signposts.

Funds and Partners applies the same logic to calls, awards, consortia and collaborators. The network brings institutions together, but a project needs legitimate partners, funding and governance. RedCLARA uses its regional position to make opportunities visible and to facilitate cross‑border teams.

This function seems small next to a submarine cable, but unknown or inaccessible infrastructure is functionally non‑existent. The failure can be at the interface: nobody knows whom to write, what requirements apply or how to get an account. RedCLARA reduces that friction, although it depends on the rules, timelines and capacities of external bodies.

To become a lasting service, hand‑offs must be explicit. The user must know when RedCLARA directly provides the resource, when it acts as a trusted intermediary and when it merely points to an option. Clarity avoids turning valuable coordination into an implicit guarantee that cannot be kept.

LA Referencia, SCALAC and the shared knowledge‑and‑compute chain

LA Referencia demonstrates that open science depends on repositories and metadata as well as on routes. Outputs are deposited in institutional or national infrastructures, aggregated regionally and made discoverable in wider systems. RedCLARA contributes connectivity, coordination and international cooperation, including OpenAIRE, but LA Referencia keeps its own governance.

A fast route does not fix a poorly described dataset. Persistent identifiers, metadata quality, licences, repository policy and preservation determine reuse. The network carries the entities; the repositories give them context, stability and visibility. Cooperation between separate entities is more important than centralised ownership.

SCALAC applies a similar logic to advanced computing. The region does not have a single RedCLARA supercomputer, but facilities governed by different countries and institutions, with different schedulers, software, queues, storage and policies. SCALAC coordinates the HPC community, and RedCLARA provides connectivity, programmes and experimental access.

An HPC federation is for prototyping, benchmarking, training and workflow development. It gives options to countries that cannot fund every class of machine. But it needs authentication, data staging, software environments, support, privacy and cost recovery. Removing the network bottleneck does not automatically turn several centres into a homogeneous system.

The strategic benefit is optionality. A team can learn on a regional resource before requesting a larger allocation; a centre can share capacity; experts can harmonise practices. The risk is that access remains episodic and dependent on projects or personal relationships, rather than a predictable service with published criteria, capacity and support.

Health, Earth observation and communities that give the network purpose

RUTE‑ALC brings together university telehealth and digital‑health communities. It connects specialists, educators and institutions that face similar problems in different national systems. Uses can include specialist consultation, medical education, image exchange, standards work, emergency collaboration and digital‑health governance.

The network can provide reliable paths, identity and collaboration tools. It does not thereby become a healthcare provider. Clinical privacy, device regulation, professional liability and consent belong to institutions and authorities. A high‑capacity connection can enable telemedicine, but it does not constitute a complete clinical service.

Copernicus Academy LAC creates another bridge. European Earth observation data can support agriculture, disasters, climate adaptation, oceans, water, forests, cities and energy. RedCLARA helps move the data and organise training, hackathons and communities. The limit is often local analytical capacity, not just image access.

TICAL plays a complementary role by bringing together higher‑education technology leaders. The 2024 edition in Rio gathered more than 200 entities according to the annual report. A conference is not physical infrastructure, but it can align network, cloud, security, identity and service decisions among organisations that must later interoperate.

These communities explain why an academic network is not simply a discount operator. It builds recurring relationships among people who run similar institutions. The trust built in commissions, events and groups can accelerate incident response or project formation in a way that a bandwidth contract does not.

Evidence must remain bounded: attendance does not prove implementation, a workshop does not create national capacity and a connection does not prove a medical or climate outcome. RedCLARA enables and coordinates conditions; it does not own every result.

EduLACSeg and the regional trust network for cybersecurity

Security coordination is a natural extension of a federated network because universities in several countries can suffer the same campaigns. eduLACSeg offers a trust framework for sharing indicators, context and expertise. The value arises when a threat seen once can be warned across several countries before each institution discovers it alone.

A July 2026 case describes information originating at CERN and distributed by RedCLARA to RNP, REUNA, RENATA, RedCONARE, RAU, CUDI and CEDIA. It is concrete evidence of exchange among seven NRENs. RedCLARA calls eduLACSeg the largest academic cybersecurity ecosystem in Latin America, but without an independent comparison the superlative must remain attributed.

Sharing threats is not automatically accurate or secure. Indicators can be outdated or context‑dependent; attribution can be uncertain; privacy laws differ; CSIRT maturity is uneven. A regional alert demonstrates distribution, not uniform response.

The 2024 report records training with LAC4 for 85 entities from almost 60 organisations in 18 countries, 30 % women, and a Security Baseline Bootcamp with RNP and GÉANT during TICAL2024. These are capacity‑building signals, not proof that every institution has implemented the same controls.

RedCLARA’s own systems also need protection: routing sessions, identity metadata, platforms and testbeds present different surfaces. Public information does not give a current, complete policy on RPKI validation, ROA coverage, IRR filters, MANRS or leak response. The next step is turning trust into measurable resilience: faster detection, coordinated exercises, common baselines and post‑incident learning.

SUBMERSE and the network as a scientific instrument

SUBMERSE extends RedCLARA’s role beyond moving data. The project studied how operating telecoms fibre can act as a sensor. Distributed Acoustic Sensing measures changes in optical backscatter, and polarisation‑state techniques observe other disturbances. The signals can help detect earthquakes, wave action, ships and environmental phenomena.

The concept changes the value of a cable. A submarine system is usually justified by capacity and resilience. Sensing adds a scientific product without building a dedicated sensor along the whole route. The same infrastructure carries data and generates data about the physical environment.

RedCLARA is relevant because of its relationship with BELLA and the Latin American end of the direct connection. It can link regional infrastructure and users with a European consortium and facilitate cable access. It did not develop every algorithm, operate every site or own all of EllaLink. The project had 24 partners and the achievements are collective.

The main phase reached a final event in February 2026. SUBMERSE should not be described as an indefinitely open programme. Its results and possible continuations remain important, but the closure of the funded phase is part of the record.

Sensing raises operational questions: the experiment must not degrade communications; data can be massive; station and fibre access need permission; some signals may be sensitive; and science needs calibration, custody and sustained analysis. For RedCLARA, the strategic value is showing that long‑term network relationships can create scientific options beyond connectivity.

An ecosystem of partners, not a collection of subsidiaries

RedCLARA’s relationship map includes NRENs, the European Commission, GÉANT, Internet2, CANARIE, development banks, EllaLink, operators, SCALAC, LA Referencia, RUTE‑ALC, LNet, Copernicus Academy LAC and SUBMERSE. The breadth can create the wrong impression that the organisation owns every initiative around it.

The evidence shows otherwise. Many entities have separate governance and mandate. LA Referencia runs repositories; SCALAC coordinates HPC; RUTE‑ALC is a health community; LNet and LACChain‑related structures evolve with their own governance; EllaLink is commercial. RedCLARA contributes connectivity, coordination, project administration, convening and sometimes services, but partnership does not mean ownership.

The distributed structure is a strength. It keeps knowledge where it resides and prevents a regional body from trying to be a universal provider. It also complicates accountability. A user can enter a RedCLARA‑branded programme whose real resource belongs to a national centre, a supplier or a partner, with different rules.

RedCLARA’s value lies in the interfaces it creates. It makes a collection of networks, labs, repositories and funding sources navigable. Quality depends on clear roles, transparent hand‑offs and honest attribution. Over‑claiming ownership would weaken the cooperative model.

Finances based on membership, projects and capacity rights

RedCLARA has no verified shareholders, market valuation or distributable profit. The statutes allow membership fees and other resources, and incorporation documents indicate that NRENs pay annual dues for governance rights, services and projects. The current amount is not public.

The large visible numbers belong to programmes, not to the organisation’s annual accounts. ALICE had €12.5 million; ALICE2 an €18 million contract; BELLA‑S1 a €12.4447 million action with €10 million from Europe; BELLA II a €28 million plan with €13 million from the Commission. They cannot be added together or called revenue: they span different years, partners, assets and in‑kind contributions.

Some value does not flow as cash through RedCLARA. An NREN can contribute fibre, equipment, people or colocation; an operator can swap capacity or co‑invest; a project can buy spectrum that is useful for years. The economic model mixes fees, grants, services, agreements and shared assets.

It has produced more than two decades of continuity, strategic rights and repeated eligibility for European programmes. It also exposes to funding cycles, national co‑financing, currencies, operator contracts and unequal ability to pay. No current audited financial statements, balance sheet, reserves or donor concentration were found. The 2024 report documents activities, but it does not substitute for accounts.

Financial transparency is a central gap. Members and partners need to know which costs are recurrent, which funds are restricted, which reserves protect operations and which suppliers concentrate risk. Long‑term infrastructure cannot be judged only by the approval of another project.

The last mile is national, institutional and unequal

A continental backbone can be impressive and distribute benefits unevenly. Countries differ in domestic fibre, NREN funding, state support, personnel, cloud access, identity and the quality of the university last mile. An institution can be inside a connected country and still be unable to use an advanced service at the required throughput or assurance level.

BELLA II exists partly because the first phase did not benefit every part of the region equally. Peru and several Central American countries need lasting routes and, in some cases, stronger national organisations. The Caribbean adds submarine costs, multiple jurisdictions and natural disasters. Building the network means strengthening NRENs, not just laying fibre.

An incentive problem arises. Strong members contribute more capacity, personnel and governance; weak ecosystems need more support. A purely usage‑based model would exclude those who most need the platform. A purely political model could ignore real costs. RedCLARA must combine solidarity, measurable obligations and a path towards stable membership.

Success is not measured by the maximum of the ring, but by how many institutions can authenticate, move data, reach compute, receive security support and take part in projects. The outcome depends on national policy and local investment outside regional control. RedCLARA can coordinate, train and negotiate; it cannot create mature domestic infrastructure by declaration.

Documentary debt hides operational maturity

The public record is rich in history and projects, but it contains conflicting metrics. Seven members, eleven connected countries and an old thirteen appear in different contexts. Current pages talk of 100, 200 and 300 Gbit/s and keep a 10G text. Manta is the current Ecuadorian node while old maps show Guayaquil. BELLA speaks of forty channels and a later proposal of forty‑five. Traffic and availability have no date or method.

Some differences are legitimate. Legal members are not the same as connected countries. A 6,000 kilometre submarine segment is not a 35,830 kilometre footprint. A grant action is not a consolidated budget. Other conflicts are outdated content that should have been removed.

Documentary debt has consequences. Governments, funders and institutions need reliable evidence to plan capacity and risk. A regional infrastructure should publish dated topology, capacity, traffic, incidents and service definitions that are comparable across years. Without them, it is unclear whether a change is growth or a change in counting.

The lack of SLA methodology, incident database, audited accounts, asset inventory and verified headcount does not deny the network. It limits conclusions about resilience, efficiency and sustainability. Transparency must be treated as infrastructure: a common layer of evidence so that members can decide on the same facts.

RedCLARA’s achievement is federation at infrastructure scale

The central achievement is not a cable, router or project, but a lasting mechanism for autonomous networks to pool demand, operate shared capacity, relate to global peers and create services above transport. ALICE created the institution and the first network. ALICE2 transferred coordination. BELLA changed the geography. BELLA II tests extension to new countries, services and workflows.

Strengths accumulate: more than twenty years of operation, direct route to Europe, 100G‑class regional capacity, NREN governance, global access and a portfolio of identity, open science, HPC, health, Earth and security. So do weaknesses: dependence on grants, national inequality, narrow legal membership, commercial suppliers and incomplete operational or financial evidence.

RedCLARA is neither a heroic self‑sufficient continental operator nor a simple project office. It is infrastructure assembled through cooperation. Its physical network depends on commercial systems, but its public value comes from using them under a regional mission and from building relationships that an ordinary circuit does not offer.

The long‑term test is to make the federation outlast the projects. It requires recurrent commitment, transparent operations, independent national capacity and clear decisions about which services deserve permanent support. If it succeeds, the region will have not only routes to global science, but an institution able to articulate routes, identities, data and computing.

Indicators for whether BELLA II becomes lasting infrastructure

The first signal is commissioning. Memoranda and requirements must be followed by identified suppliers, accepted routes, real traffic and protected capacity for Peru and Central America. Reports must separate initial minimum, lit capacity, protected aggregate and usable bandwidth. The fifteen‑year aim matters only if operation, maintenance and upgrades are contractually sustainable.

The second signal is membership. RedCLARA has seven legal NRENs and eleven connected countries. It is worth watching whether target countries become lasting members, partners or direct entities, and how each status affects voting, fees, service rights and obligations. A larger network without a larger recurring base would increase burden and keep control concentrated.

The third signal is operational evidence: dated topology and traffic, SLA methodology, incidents, route‑security policy, RPKI posture and measured diversity between EllaLink, Miami and alternatives. A major fault or leak would test both physical and policy resilience.

The fourth is service conversion. BIO+IA, HPC, MiLab, identity, eduLACSeg and open science must be measured by recurring users, production assignments, support, data governance and post‑pilot funding, not just by attendees. A pilot becomes infrastructure when users can plan around it.

The final signal is financial. Audited accounts, reserves, fee concentration, supplier exposure and the ratio between unrestricted and restricted funds would improve credibility. The decisive question is maintaining capacity and services between European cycles without leaving weak NRENs as temporary beneficiaries.

Evidence‑backed scenarios

In a successful expansion scenario, Peru and Central America obtain lasting, scalable routes and the regional base widens. The European route, North American gateways and an improved Pacific presence support larger instruments and shared computing.

In a service‑federation scenario, connectivity becomes less visible because identity, cybersecurity, data, HPC and testbeds are the daily products. Value increases, but so do software, privacy, support and recurring funding obligations.

In a concentrated‑governance scenario, membership stays at seven while projects and connections grow. The organisation is regional in operation but depends on a small core for control and cost, raising representation and succession questions.

In a project‑cycle‑dominated scenario, innovations are launched with grants and struggle to survive. The backbone continues, but higher‑level services oscillate with programmes. This is the main risk if no permanent model emerges.

Co‑investment with operators can improve sustainability through swaps, joint construction and long rights. It also increases dependence on a limited number of suppliers and demands strong maintenance, upgrade and exit clauses.

A network‑as‑instrument scenario would extend SUBMERSE to more regional fibre and generate environmental and geophysical data. It remains prospective, but it would give a distinctive scientific role that commercial capacity does not provide.

Who controls the system and who bears the risk

Control is distributed. NRENs govern the association through the Assembly and Board. The Directorate and technical groups control regional implementation. National networks control domestic infrastructure and institutional access. GÉANT and other peers control their systems. Funders influence through programmes. Operators, cable companies, colocations and manufacturers control physical dependencies. Scientific communities create demand but do not always have direct regional governance rights.

The structure prevents any single actor from owning everything, but it separates decision rights from consequences. A Board can approve a programme whose last mile is paid by an NREN; a funder defines four years and members inherit decades of maintenance; an operator owns the only practical route; a user depends on a service without knowing who can fix it. These interfaces are the primary control surface.

Incentives must recognise lasting contributions, not just project participation. Fees, fibre, personnel, operations, data custody and security response are distinct forms of value. Countries receiving expansion need a route to national ownership and recurring obligations. Strong members need to know that solidarity does not become an unlimited cost transfer without accountability.

Contracts must preserve diversity, upgrade rights, auditability, exit and continuity under supplier change. Spectrum rights reduce price risk, but they can become stranded if stations, equipment or maintenance fail. The critical decisions are not only how much to buy, but which dependencies will be hard to unwind.

Decisions that cannot wait until the next grant

First, RedCLARA must decide whether it remains primarily a connectivity federation or becomes a permanent platform operator. MiLab, identity, cybersecurity and AI add value but also data, software, staff and liability. Permanent services need clear ownership, lifecycle funding, incident response and retirement rules.

Second, it must resolve the gap between legal membership and operational reach. Connected countries need status, voice and a path to membership or another accountable arrangement. Keeping the gap informal eases short‑term expansion but creates ambiguity about who governs, pays and can leave.

Third, financial transparency must become a governance tool. Accounts, reserves, fees, restricted funds and supplier concentration determine whether the network survives a disruption or emergency. Members cannot commit for the long term without evidence.

Fourth, RedCLARA should publish a continuity architecture: route diversity, incident escalation, configuration backups and logs, identity and platform recovery, alternatives under cut and succession for project‑funded services. Essential functions must survive the failure of a grant, supplier, route or national institution.

The irreversible risk is not just delay. It is accumulating critical services and long dependencies faster than recurring authority, funding and operational evidence. The irreversible opportunity is to use BELLA rights, NREN governance and global partners to create a regional common good that no single country could build alone.

The long‑term meaning of a shared regional backbone

RedCLARA demonstrates a form of infrastructure sovereignty based on cooperation, not isolation. Latin American science gains agency not because the organisation owns every fibre, but because national networks jointly control a regional layer, hold long‑term strategic capacity and participate in global partnerships on more balanced terms. Commercial dependence remains, but mission, routes and upgrade decisions are not left entirely to the commodity Internet market.

The model matters beyond research. It teaches how unequal institutions can aggregate demand without giving up national autonomy and how grants can turn into rights and organisations that outlive the project. It also shows the limit of federation: coordination does not indefinitely compensate for weak domestic networks, opaque finances or ambiguous accountability.

The decisive choice is whether the next phase will be expansion of a map or maturation of an institution. More countries and more speed matter, but the lasting asset is the system of governance, operations, identity, data, security and scientific cooperation. If it becomes transparent, resilient and genuinely regional, RedCLARA will remain one of Latin America’s most important shared digital infrastructures. If not, the traffic can continue while its higher ambitions depend on the next external programme.