RedCLARA: Building Latin America's Research Backbone Through Federation, Not Full Ownership

Summary

• RedCLARA is a non-profit association registered in Uruguay, uniting national research and education networks; it is neither a shareholder-owned company nor an operator controlling access to each campus
• Eight major routing nodes and long-term international capacity form a 100G-class backbone; public materials list 100, 200, and 300 Gbps links, as well as a smaller shared Central American segment
• ALICE and ALICE2 established the institution and the first network; BELLA obtained long-term capacity on the direct Portugal–Brazil route, reducing reliance on indirect paths via North America
• The test of BELLA II is converting project funding into infrastructure sustained by member support, while strengthening weaker national networks, improving public evidence, and avoiding bottlenecks from suppliers or membership structures

BELLA Changed Network Geography

BELLA (Building the Europe Link to Latin America) combines submarine and terrestrial components. BELLA-S obtained long-term spectrum on the direct submarine cable between Portugal and Brazil, while BELLA-T upgrades the South American terrestrial backbone. The actual cable system is EllaLink, approximately 6,000 km from Sines to Fortaleza.

The project configured two 100 Gbit/s links: one connecting GÉANT and RedCLARA, and another supporting Copernicus Earth observation-related needs. Research and education traffic has been using the direct path since 2021. Project materials mention latency reductions of up to 50% for some paths, a figure that must be qualified for specific source-destination pairs and cannot be generalised.

EU CORDIS records the BELLA-S1 action cost at €12.4447 million, with €10 million from the EU; the broader BELLA also includes terrestrial works, other EU departments, and national in-kind contributions. Different totals may correspond to different scopes and cannot simply be summed.

A 2024 report states the broader BELLA infrastructure as 35,830 km and Nx100G. This figure includes multiple submarine and terrestrial components and must not be misstated as the length of the single EllaLink cable.

Forty optical channels do not mean owning the submarine cable

BELLA has obtained long-term rights to forty optical channels on the EllaLink system. This is one of RedCLARA's most important infrastructure facts and also the most easily exaggerated. Spectrum rights give the research community dedicated, upgradeable, long-term capacity, but do not mean exclusive ownership of the entire submarine system.

EllaLink is a commercial submarine cable with its own investors, operators, landing stations, and other customers. RedCLARA and GÉANT control the contracted capacity rights but do not unilaterally control cable repair ships, all repeaters, the physical route, every landing station, or other traffic on the cable.

The value of the forty channels lies in optionality. The alliance is not locked into a single initial fixed bandwidth; it can light or upgrade capacity with newer-generation optical systems as contract and equipment conditions permit. This reduces the risk that early procurement speeds become quickly obsolete.

Ownership boundaries also determine fault responsibility. An incident may require the cable operator, landing stations, optical equipment vendors, RedCLARA, GÉANT, or national NRENs to respond. True resilience comes from contracts, upgrade mechanisms, and alternative routes, not from a label on a map that says “direct.”

BELLA and the South Atlantic Direct Route

The most visible change BELLA brings to Europe–Latin America research connectivity is geographic. Previously, traffic could route via North America even when both ends sat on opposite shores of the South Atlantic. The direct Sines–Fortaleza route provides a shorter alternative for many paths and leaves a long-term upgradeable asset for the research community.

It does not guarantee that latency is halved for all paths. Actual gains depend on source, destination, domestic access, congestion, and backhaul. The project’s claim of “up to a 50% reduction” should be qualified to specific paths. A more robust conclusion is that European connectivity no longer depends entirely on North American gateways, and routing diversity has improved.

Large-scale data transfers, astronomy, Earth observation, distributed computing, academic telemedicine, and interactive collaboration can all benefit from more direct, more predictable paths. RedCLARA and GÉANT have gained a more controllable research interconnection capability than ordinary public internet transit.

Direct submarine cables are still subject to cable cuts, power supply failures, landing station outages, optical equipment failures, and repair faults. Resilience requires alternative paths, failover testing, and sufficient capacity via Miami or other directions when EllaLink is unavailable. “Direct” improves architecture but cannot replace redundancy.

BELLA is also a long-term institutional collaboration. The capacity was obtained through a research cooperation programme, not a standard commercial internet procurement. This gives the research community greater influence over use and upgrades, but commercial operators remain a physical dependency that cannot be ignored.

Spectrum Rights, Terrestrial Coverage, and Ownership Boundaries

The phrase “RedCLARA network” can obscure many forms of control. The organisation may own routers, rent rack space and power, purchase managed services, hold IRUs, obtain spectrum rights in third-party cables, or exchange capacity with partners. Each form carries different durations, maintenance, upgrade rights, and risk profiles.

The marine segment uses BELLA’s long-term spectrum on EllaLink; the terrestrial segment is assembled by BELLA-T and members combining domestic fibre, wavelengths, and services to reach PoPs and institutions. The 35,830 km figure in the 2024 report is a project-level coverage concept and must not be stated as the length of fibre fully owned by RedCLARA.

Distinguishing ownership avoids two opposite errors: underestimating the strategic value of long-term spectrum rights, and exaggerating RedCLARA’s control over construction, landing stations, cable repairs, optical equipment, and domestic contracts.

True robustness lies in the contract terms: how long the rights last, who is responsible for maintenance, whether they are upgradeable, what happens if the operator is acquired or bankrupt, whether access to landing stations is guaranteed, and what exit and alternative routes exist. A line on a map has full meaning only when these questions are answered.

The federal model allows the organisation to achieve regional coverage without owning every physical asset, but at the cost of greater supplier governance requirements. It must be clear which dependencies are replaceable and which have become hard-to-reverse single points of failure.

A Continental Network, But Not One Controlled by a Single Operator

Simply calling RedCLARA “the Latin American research network” easily creates the misunderstanding that it controls everything from the international backbone down to the university port. The actual structure is federal. Researchers typically traverse their campus or institutional network, then their national research and education network (NREN), and only then use the regional and intercontinental layers provided by RedCLARA.

Brazil’s RNP, Chile’s REUNA, Colombia’s RENATA, Ecuador’s CEDIA, Mexico’s CUDI, Costa Rica’s RedCONARE, and Uruguay’s RAU retain their own national governance, domestic infrastructure, institutional relationships, and operational responsibilities. RedCLARA does not replace these networks; it connects them, aggregates common needs, coordinates capacity and services, and links to global research networks such as GÉANT, Internet2, and CANARIE.

An end-to-end research path may pass sequentially through a campus LAN, an institutional firewall, a domestic access line, the NREN backbone, RedCLARA, international peering, the destination country’s NREN, and the remote experimental facility. A regional 100 Gbit/s link cannot compensate for campus egress congestion, limited public evidence domestic lines, storage misconfiguration, or destination-end policy restrictions; conversely, a single national network cannot independently negotiate all intercontinental routes, identity trust relationships, and collaborative services.

The most precise definition is therefore a “shared coordination and infrastructure layer.” Sharing here refers to governance, engineering, long-term capacity rights, and community services; it does not mean owning every fibre segment, every router, every PoP, or submarine cable on the map. RedCLARA’s core mechanism is organising multiple national systems and commercial infrastructure contracts into a regional network with a public research mission.

Why Latin America Needs a Regional Layer

The fundamental problem at inception was fragmentation. In the early 2000s, some countries already had or were building national research networks, but the region lacked a long-term continental institution capable of aggregating demand, operating a regional backbone, and negotiating on equal terms with European and North American research networks. International leased lines were expensive, national telecom markets varied widely, and commercial routing served operator economics first rather than distributed scientific collaboration.

A university with good domestic connectivity could still depend on expensive, circuitous international transit when collaborating with an institution in another Latin American country. Traffic between Europe and Latin America often routed via North America because existing cable and interconnection structures were easier to procure. The cost was not only latency but also weak bargaining power, limited route diversity, and critical service priorities set by actors outside the research community.

Creating a single operator to centrally control the continent was not realistic. Countries differed enormously in NREN maturity, public funding, regulation, fibre resources, university systems, and international landing points. National organisations had to retain legal and operational autonomy. The feasible solution was therefore federal: domestic networks handle national responsibilities, while a regional association manages the common layer and interfaces with global partners.

This design also creates institutional value. A single NREN struggles to influence large international research programmes or provide sufficient demand justification for dedicated routes. A regional organisation can pool research scenarios and funding, combine grants with national assets, and act as a unified counterparty towards the European Commission, GÉANT, development banks, operators, and cable companies.

Toledo, Valle de Bravo, and the Legal Association

The institutional concept took shape in 2002. Representatives of Latin American NRENs discussed a regional network at a CAESAR feasibility meeting in Toledo, Spain, and continued in Rio de Janeiro, Buenos Aires, Santiago, and Mexico. The discussions covered not only topology but also membership, the relationship between a regional institution and national networks, decision-making processes, and how externally funded projects could become a sustainable organisation.

The legal timeline has three dates that should all be retained. RedCLARA records show the statutes were signed on 9 June 2003 in Valle de Bravo, Mexico; the OAS civil society record uses 10 June as the date of establishment; RedCLARA also states that its legal existence was recognised by Uruguay on 23 December 2003. They can be understood respectively as the founding agreement, an administrative record, and subsequent legal recognition.

The safest legal description is: a non-profit international civil association with headquarters in Montevideo, recognised under Uruguayan law. RedCLARA sometimes uses the English term “international law organisation,” but the statutes and registration materials available do not establish that it was created by international treaty or that it has intergovernmental organisation status. It is a membership association, not a supranational regulator or government body.

The current statutes were adopted in November 2019 and published in January 2020. The Assembly is the highest governance body; the Board sets direction; the Finance Committee oversees; the Technical Committee brings together NREN experts; and the Executive Directorate handles implementation. The network’s legitimacy derives from member collaboration, not shareholding or territorial authority.

Seven Legal Members, Eleven Connected Countries, and a Larger Mission

The current public membership directory lists seven NRENs: RNP, RENATA, RedCONARE, REUNA, CEDIA, CUDI, and RAU; yet the latest advanced connectivity page states that the network connects eleven countries: Argentina, Brazil, Chile, Colombia, Costa Rica, Ecuador, El Salvador, Guatemala, Mexico, Paraguay, and Uruguay. Older materials also mention thirteen countries. These three figures must not be conflated.

Legal members determine votes, obligations, and association governance; operational connectivity can exist through partnerships, transitional arrangements, or project presence, even if the relevant national network is not currently a full member. Historical numbers may reflect past members or old project scopes. It should therefore be stated clearly as “seven current legal NREN members and eleven currently connected countries.”

This gap creates strategic tension. RedCLARA serves a region larger than the association that formally governs it. A country can obtain connectivity or participate in BELLA II without an equivalent seat in the Assembly, while the seven members bear more governance, technical work, and likely recurring financial responsibilities. This does not automatically imply institutional failure, but it affects representation, cost-sharing, and the long-term legitimacy of the regional platform.

The 2025–2027 Board comprises José Palacios of REUNA as President, Eduardo Grizendi of RNP as Vice-President, Juan Pablo Carvallo of CEDIA as Treasurer, Carlos Gamboa of RedCONARE as Secretary, and Eduardo Grampín of RAU as Director. Luis Eliécer Cadenas is the Executive Director. Public materials do not fully disclose recent Assembly minutes, voting weights, current membership fees, or how non-member connected countries participate in major decisions.

ALICE Provided Start-up Capital for the Regional Network

RedCLARA’s legal founding and its first operational network are closely tied to ALICE (America Latina Interconectada Con Europa). The European Commission and DANTE signed the contract in June 2003, with a total project of €12.5 million—€10 million from the EC and €2.5 million from Latin American partners—running, after extension, until March 2008.

ALICE simultaneously addressed start-up funding, joint procurement, operational objectives for the new association, and connection to GÉANT. It did not require an immature member revenue model to be in place first; instead, it allowed the network to prove its value through live operation while the institution consolidated.

The early network was modest by today’s standards, but it established regional PoPs, production interconnections between participating NRENs, and technical relationships with European and North American research networks. These relationships later extended into identity, middleware, grid computing, open science, and joint projects. A live network transformed the association from a mere conference platform.

ALICE also left behind the core problem of sustainability: fixed-term grants can build a network but cannot automatically pay for its operation after the project ends. Long-term survival required increasing regional contributions, reducing recurring costs, securing longer-term capacity rights, and building an institution capable of continuous negotiation and operation.

ALICE2 Shifted Coordination to Latin America

ALICE2 started in December 2008 with an EC contract of €18 million. The key institutional shift was that RedCLARA became the coordinator rather than merely a regional beneficiary, handing more of the alliance management, implementation responsibility, and leadership to Latin American institutions.

Project objectives included strengthening sustainability, broadening national participation, reducing recurring costs through dark fibre, wavelengths, and IRUs, developing user communities, and training technical staff. Moving from short-term leased lines to long-term infrastructure rights was as much an economic strategy as a technical upgrade.

Different sources place the end date in March 2012 or January 2013. A safe formulation should acknowledge operational and administrative closure during 2012–2013 and note the source discrepancy. This minor contradiction illustrates that project delivery, operational end, and administrative closure cannot be silently conflated.

ALICE2 proved that RedCLARA could continue operating a production network after a single grant period, but it also showed that “regional leadership” does not equal full self-sufficiency. Expansion still relied on European collaboration, national assets, and commercial contracts.

From Short-term Leased Lines to Long-term Capacity Rights

The economic model of a regional research network depends heavily on how capacity is acquired. Short-term managed leased lines provide defined services, but the operator faces renewals, price increases, and changes in carrier strategy. Dark fibre, spectrum, IRUs, or capacity exchanges require more upfront engineering but can offer longer-term control and lower marginal upgrade costs.

RedCLARA’s actual network consists of multiple rights types: it may own routers, rent rack space and power, purchase operation and maintenance services, hold spectrum rights in cables it does not own, or exchange capacity with partners. The logical network is therefore an assembly of technical assets, contractual rights, and third-party infrastructure.

Long-term rights do not eliminate risk. They depend on cable lifespan, landing station access, terminal equipment, submarine or terrestrial maintenance, counterparty solvency, and succession clauses. Spectrum rights without funding for lighting, upgrading, and maintenance can become stranded assets.

The advantage lies in aligning infrastructure horizons with research collaboration horizons. Research programmes and NREN relationships often span decades; if only annual connections are bought, the network remains exposed to price and contract volatility. Converting part of a grant into long-term usable rights is key to retaining value after the project ends.

Before BELLA, Routes to Europe Often Passed Through North America

The absence of a direct, high-capacity research path between Europe and Latin America affected performance and dependencies. Even when both ends were on opposite shores of the South Atlantic, traffic could still enter a North American node and then cross the North Atlantic, because such commercial routes were easier to procure.

The detour added latency, concentrated failure domains, and embedded European research relationships in a topology shaped by North American hubs. For interactive science, Earth observation, large-scale data transfers, and remote instrumentation, distance and jitter matter alike.

These paths were not technically “wrong” but a consequence of cable and interconnection economics. Changing it required investment in submarine and terrestrial capacity, not merely altering routing preferences. The research community had to aggregate demand to a scale that could support a direct long-term rights commitment.

BELLA’s goal was therefore not just more bandwidth, but physical and institutional diversity: a direct South Atlantic path that the research community could use and upgrade over the long term.

Current Topology: Eight Major Nodes, Diverse Capacity Forms

RedCLARA’s latest technical description lists eight major routing nodes: Buenos Aires, Fortaleza, Porto Alegre, São Paulo, Santiago, Manta, Panama City, and Miami. South American NRENs connect to these nodes and interconnect via rings at 100 Gbit/s or higher; the shared Central American infrastructure is described as 20 Gbit/s. Fortaleza is the main direct entry point to GÉANT and research partners in Europe and Africa, while Miami is the primary gateway to Internet2, CANARIE, and North American partners.

The advanced connectivity page also mentions 100, 200, and 300 Gbit/s links, eleven countries, and more than 1,300 institutions. The same site still retains old 10 Gbit/s language. Newer topology and project materials support a “100G-class network” assessment; the 10G phrasing appears more like documentation debt than evidence of current capability.

These rates do not mean that available capacity is identical on every path. Line rate, protected aggregate capacity, reserved resources, and the bandwidth actually delivered to members are different concepts. Logical rings may also share the same commercial optical cable or physical corridor, so logical redundancy does not equal end-to-end physical diversity.

Manta provides a Pacific-side node, Fortaleza anchors the South Atlantic, and Panama City and Miami serve Central American, Caribbean, and North American directions. The topology is shaped jointly by cable landing points, domestic markets, NREN collaboration, and existing fibre; it is not a perfectly symmetrical continental ring.

“More than 1,300 institutions” is a coverage metric, but no public methodology explains how many are active, at what speeds, or which regional services they actually use. Maturity should be measured by capacity, traffic, availability, services, and repeat usage, not by a single total.

AS27750, Peering, and Traffic Class Separation

AS27750 is RedCLARA’s public autonomous system number. Through it, the organisation carries or exchanges routes with member networks, research and education peering, and international services. Public databases show relationships with GÉANT, ESnet, Internet2, CANARIE, RNP, REUNA, RENATA, RedCONARE, CEDIA, and TENET, among others, but these databases are snapshots in time and cannot be taken as a complete contractual topology.

A research network is not “a faster normal internet.” Eligible research and education traffic can reach NRENs, research facilities, and international partners via dedicated or private paths, avoiding general commercial transit when peering relationships exist. This can lower costs, improve predictability, and reserve capacity for data-intensive applications.

RedCLARA simultaneously provides separation of commercial internet traffic, peering, and global connectivity. Universities need both access to research resources and the public internet; the two traffic classes may use different routing policies, providers, and settlement models. The research network is not physically isolated from the internet but achieves logical distinction through eligibility, routing, capacity, and governance.

Selective peering means direct connections are established only where there is mutual value; it does not guarantee that any network can connect. Even with colocation in Miami or São Paulo, ports, cross-connects, route filtering, policies, and ongoing operations are still required.

Public materials do not fully document current RPKI validation, ROA coverage, IRR filtering, route leak protection, BGP communities, and per-peer availability. For a regionally critical network, clearer routing security disclosures—without overexposing sensitive information—would significantly improve assessability.

Layer-2 Circuits, Operational Collaboration, and Regional Responsibility Boundaries

RedCLARA’s services include dedicated Layer-2 virtual circuits. VLANs can make laboratories, data centres, or scientific instruments in different countries appear on the same Ethernet segment, suitable for large data transfers, distributed experiments, or specialised workflows that are hard to satisfy with normal routed IP.

A Layer-2 service typically spans multiple administrative domains. The institution provides the local port; the national NREN brings traffic to the regional network; RedCLARA coordinates the international segment; and the destination reverses the process. VLAN, MTU, redundancy, bandwidth, addressing, security, and change windows must all align across the parties.

This means operational responsibility is distributed. A fault may occur at the campus, the NREN, RedCLARA, a transit provider, or the remote institution. The regional NOC needs clear circuit identifiers, 24/7 contact details, demarcation points, and escalation paths; otherwise users only see “the network is down” without knowing who has the authority to fix it.

RedCLARA publicly mentions engineering, NOC, and security teams but does not fully publish SLA calculation methods, historical repair times, or an independent incident database. Older pages report figures like “100% average monthly availability” and “7.7 PB annual traffic,” but without scope and methodology, these cannot be treated as audited results.

The implication of the federal structure is that RedCLARA is responsible for the regional segment and coordinates NRENs, but it cannot unilaterally guarantee every component. Professional services require that users understand responsibility boundaries and that the federation becomes a chain of collaborative support, not a chain for shifting blame.

BELLA II: Extending Through Co-investment, Not a Finished Map

RedCLARA and the European Commission signed BELLA II on 14 December 2022. The programme runs for 48 months, with a total envelope of €28 million, of which the EC commits €13 million; RedCLARA must mobilise at least €15 million from governments, development agencies, enterprises, banks, and other partners.

Infrastructure objectives include connecting or strengthening Peru, Costa Rica, El Salvador, Guatemala, and Honduras, with possible extension to Mexico, the Caribbean, Paraguay, Uruguay, Bolivia, and Belize. These country names represent objectives, dialogue, or potential scope; they must not be drawn as a network in production on 3 August 2026.

A 2025 competitive dialogue seeks fibre systems with at least a 15-year lifespan, an initial minimum of 20 Gbit/s in Central America and 100 Gbit/s in Peru, and scalability. Solutions may come from existing fibre, new builds, capacity exchanges, co-investment, or hybrid models.

Flexible procurement helps attract funding but complicates accountability: who builds, who owns, who maintains, how much capacity is reserved for research, how countries join, and who carries operations after the project must all be clear. A 15-year capacity right only truly makes sense if power, equipment, maintenance, and upgrades are also funded over the long term.

BELLA II should therefore be measured by accepted routes, enforceable rights, actual production traffic, and post-project operating arrangements. Memoranda of understanding and policy dialogue represent progress, but they are not live services. Rigorously distinguishing announcements from production is essential for maintaining project credibility.

From Transport Network to Regional Research Platform

The first-generation RedCLARA proved that national research networks could exchange traffic over a regional backbone. The subsequent question was: what common capabilities can be built on top of that backbone? The portfolio has since expanded to identity, file transfer, collaboration environments, training, HPC, open science, cybersecurity, and thematic communities.

This is not a departure from the network but an increase in its value. High-capacity paths have limited research impact without identity authentication, compute resource discovery, large file movement, partner matching, and cross-border support. The application layer and communities turn raw bandwidth into usable infrastructure.

BELLA II institutionalises this platform logic through innovation centres, testbeds, open calls, training, strategic dialogues, and funding partner matchmaking. The 2024 Agrifood Ideathon involved 125 experts from 28 countries, and early HPC testbeds have brought training and prototyping to shared compute resources.

Testbeds differ from production services. They are typically aimed at selected teams, with limited capacity, expert support, and an experimental purpose. They can expose technical and organisational barriers before long-term builds, but they can also mislead when a successful demonstration is packaged as general availability.

RedCLARA’s opportunity is to become a federation of services and testbeds, not just a transport network. At the same time, identity, storage, compute, notebooks, collaboration, and sensitive data all add security, privacy, support, and lifecycle responsibilities. The underlying resources remain operated by NRENs, SCALAC members, universities, clouds, and project partners; RedCLARA’s role is to integrate, curate, coordinate, and provide a regional access point.

BIO+IA: Early Adopter Evidence Should Keep Its Boundaries

BIO+IA Early Adopters is one of the clearest examples of platformisation. The programme opened in 2026 under BELLA II, selecting five projects from six countries and providing them with training, expert support, and an AI-bioinformatics test environment. Topics range from breast cancer biology, molecular screening, antimicrobial or anti-mercury mechanisms, drug repurposing for liver disease, and translational medicine.

At the July 2026 Demo Day, teams reported that the environment helped them search, integrate, and interpret biological information. This is valid evidence of a supported research workflow, showing that networking, specialised tools, and expert guidance can be combined at the regional level.

It does not constitute clinical efficacy evidence. A model finding a candidate molecule does not mean the treatment is effective, and faster analysis does not equate to completed experimental validation, regulatory review, or medical deployment. The programme serves a limited set of projects, and results are reported by entities and organisers, not by an independent clinical trial.

These qualifications do not diminish the programme; they clarify the right unit of achievement: RedCLARA supported a structured multinational experiment, lowered the barriers for participating teams to access tools, data, and know‑how, and built multinational research relationships.

The next test is continuity. A demonstration becomes infrastructure only when teams can return, new groups can join, data governance is clear, compute and storage are funded, and successful workflows can migrate into national or institutional systems. Repeat usage, publications, independent evaluations, and post-pilot funding are key to judging whether BIO+IA can become a long-term service.

The broader measurement principle is to see how far participation events are from long-term capability. An Ideathon can count entities; a testbed can count experiments; real infrastructure measures persistently maintained environments, repeat users, governed data, trained operational staff, and projects that continue after the grant or event ends.

Identity Is Also Infrastructure: FIEL, eduGAIN, and eduroam

No matter how good the network path, it is meaningless if researchers cannot prove their identity to the destination service. FIEL supports national identity federations; eduGAIN allows participating federations to exchange metadata and lets users access international services with their home institution identity; eduroam applies a similar trust model to roaming network access.

These systems do not build a regional password database. In a typical federated login, a service redirects the user to their home institution; the institution authenticates and returns a signed assertion or selected attributes; the service then enforces its own access policy. RedCLARA strengthens national federations and connects to the global trust fabric, usually without touching all user passwords.

This structure distributes trust and risk. Universities retain account responsibility; researchers can use the same institutional identity to access repositories, compute platforms, and collaboration tools. However, assurance levels vary, metadata can become stale, attribute semantics may be inconsistent, and a compromise of an identity provider or signing key can affect multiple services.

eduGAIN addresses discovery and metadata exchange; it cannot solve all legal issues. Services still decide which institutions and assurance levels to accept; institutions must protect keys, manage account lifecycles, and respond to incidents. Cross-border use involves consent, attribute minimalisation, privacy, and legal bases.

In eduroam, the visited institution provides the network and proxies authentication requests through national or regional RADIUS hierarchies to the user’s home institution, which validates credentials. Security depends on correct client configuration, certificate validation, the home institution’s identity system, and local network controls.

Identity federation is part of infrastructure continuity: a supercomputer may be reachable at the IP layer, but if the trust chain fails, the intended user still cannot use it. By coordinating both connectivity and identity, RedCLARA addresses failure modes that a pure transport network would ignore.

MiLab, eNVIO, and the Research Collaboration Layer

MiLab puts several open-source tools behind federated access: a Dataverse‑based data environment, a GitLab-based development environment, a Mattermost communication space, and notebook computing associated with JupyterHub, Python, and R. It can support teams from communication and code development through to analysis, data curation, and publishing.

Multinational teams often stitch these functions together from unrelated services, with disparate accounts, data policies, and support models. A regional workspace can reduce fragmentation and offer a common environment for projects spread across institutions and countries. RedCLARA has recorded use cases such as geographically distributed data collaboration.

MiLab must not be characterised as a general-purpose regional cloud. It does not automatically provide large-scale HPC, unlimited storage, compliance certification for any sensitive data, or permanent legal preservation; capacity, service levels, and hosting arrangements may also change. Specific projects still need to match security, data, and performance requirements.

eNVIO addresses the transport of large files over academic infrastructure. Email and ordinary consumer cloud drives are unsuitable for many research datasets. Its real value depends on file size limits, retention, encryption, support, and national network integration—metrics that are not fully disclosed.

RedCLARA also operates a regional training platform and a learning management environment in partnership with CEDIA. Training can turn expertise into courses and certifications but cannot replace permanent positions, production experience, and institutional responsibility.

This collaboration layer is a microcosm of the federal model: RedCLARA integrates tools and access points, NRENs bring users, institutions provide identity and project environments, and researchers do the work. It works only when responsibilities are clear; a regional platform cannot unconditionally compensate for all local personnel, data governance, and application support gaps.

Ventanilla Abierta: Navigating a Fragmented Ecosystem

Regional research infrastructure, even when it exists, can be hard to discover. Researchers may know that there is HPC, storage, identity services, or specialised networking somewhere in Latin America, but not who operates it, whether their project qualifies, or where to start. Ventanilla Abierta was built precisely as a single‑point‑of‑contact advisory layer.

It does not own every resource it recommends; instead, it translates research needs into ecosystem pathways: connectivity via national NRENs, compute via SCALAC or other centres, access to data, storage, or identity services, finding international partner resources, or entering project and funding channels. Its value lies in coordination and discoverability.

This makes Ventanilla Abierta a navigation infrastructure. Physical resources are underused often not because there is no demand but because institutional interfaces are scattered. If researchers must separately understand multiple NRENs, project rules, and technical teams, regional options may be abandoned. A trusted single entry point can reduce those transaction costs.

Funds and Partners addresses a related problem: research collaboration needs not only networks but also calls for proposals, complementary institutions, and funding. Connecting technical needs with partners or funders can turn a potential idea into a real project that actually uses the infrastructure.

But RedCLARA’s remit is limited. Advice is not a reservation; a funding call is not an award; an introduction is not an alliance agreement. The service should be measured by successful transitions and sustained projects, not merely by counting how many enquiries are received.

LA Referencia, SCALAC, and the Shared Knowledge–Compute Chain

RedCLARA’s relationships in open science and high‑performance computing show how transport can become an ecosystem. LA Referencia aggregates metadata and research outputs from national repository systems; SCALAC links advanced computing institutions and promotes HPC collaboration. They are independent alliances, not RedCLARA subsidiaries, but the regional network and coordination body help them operate across borders.

The open science chain starts with institutional or national repositories. Researchers deposit papers or data according to local policies; national systems organise metadata and access; LA Referencia provides regional discovery and interoperability, and connects Latin American outputs to broader contexts through relationships such as OpenAIRE. RedCLARA provides connectivity, project coordination, and institutional ties but does not own each repository or output.

Open science needs more than bandwidth. Metadata quality, persistent identifiers, licences, preservation, repository governance, and research incentives determine whether outputs can be found and reused. High-speed networks can move data but cannot make undocumented files reproducible.

SCALAC addresses analogous issues in computing. Latin American HPC is spread across universities, national centres, and other institutions. A regional alliance can make resources and experts more visible and support testbeds, training, and multinational projects that are hard for a single country to provide.

A regional HPC federation is not a single supercomputer. Schedulers, software stacks, access policies, queues, storage, and billing differ; data must be staged, users must authenticate, allocations must be governed, and sensitive workloads may face cross‑border restrictions. The network removes one bottleneck but exposes others.

The strategic gain is optionality: countries that cannot buy every specialised compute themselves can still participate; idle facilities can gain partners; teams can prototype before applying for larger resources. The risk is that access remains permanently project-based and relationship-driven, without public, predictable rules on eligibility, capacity, and cost.

Telemedicine, Earth Observation, and Communities That Give Networks Purpose

RUTE-ALC connects regional university telemedicine and digital health communities. Potential uses include expert consultations, medical education, image exchange, standards work, emergency collaboration, and digital health governance.

A research network can provide reliable paths, identity, and collaboration tools, but that does not make it a healthcare provider. Clinical privacy, medical device regulation, professional liability, and patient consent remain with health institutions and public authorities. High-capacity connectivity is one condition for telemedicine, not a complete clinical service.

Copernicus Academy forms another bridge in Latin America. European Earth observation data can support agriculture, disaster response, climate adaptation, oceans, water, forestry, urban planning, and energy. RedCLARA helps transport data and organise training and innovation; the real bottleneck is often analytical capacity, not just raw image access.

TICAL brings together higher education technology leaders and the engineering community. The 2024 event in Rio de Janeiro reportedly had over 200 entities, according to the annual report. A conference is not physical infrastructure, but it can create alignment on procurement, security, cloud, identity, and service priorities, building relationships among institutions that will later need to interoperate.

These communities explain why a research and education network is not just a “discount operator.” Persistent trust accelerates incident collaboration and project formation. But attendance is not implementation; a workshop is not national capacity; network connectivity is not a health or climate outcome. RedCLARA’s contribution is to create and coordinate the conditions for action, not to own every result.

EduLACSeg and the Regional Cybersecurity Trust Network

Academic institutions often face similar attacks across borders; cybersecurity coordination is a natural extension of the federal network. eduLACSeg provides a trust framework for sharing indicators, context, and response experience, so that activity spotted in one place can be reported before other institutions discover it independently.

A July 2026 example shows threat information originating at CERN being distributed via RedCLARA to RNP, REUNA, RENATA, RedCONARE, RAU, CUDI, and CEDIA. This is concrete evidence that the regional layer can pass security intelligence. RedCLARA describes it as the largest academic cybersecurity ecosystem in Latin America, but an independent comparative definition is lacking, so it should be kept as the organisation’s own characterisation.

Threat sharing is not automatically accurate or safe. Indicators can be wrong, stale, or highly contextual; attribution may be uncertain; personal data and confidentiality rules differ across countries; NREN response capability can range from mature security teams to small groups with limited tools. A regional alert proves distribution, not uniform remediation.

The 2024 report records that RedCLARA and LAC4 provided training to 85 entities from nearly 60 institutions in 18 countries, with 30% women, and also held a Security Baseline Bootcamp with RNP and GÉANT at TICAL2024. These are capacity-building signals, not proof that all institutions have implemented the same controls.

RedCLARA itself must also protect routing sessions, identity metadata, collaboration platforms, and testbeds. Public materials do not provide a complete, current picture of RPKI, ROA, IRR filtering, incident disclosure, and response time methodologies. Long-term trust requires rules on intelligence classification, redistribution, data protection, attribution, and post‑incident review, as well as measurable technical controls.

SUBMERSE: Turning the Network Itself into a Scientific Instrument

SUBMERSE extends RedCLARA’s role from “transporting science data” to “the infrastructure itself generating science data.” The project explores using technologies such as distributed acoustic sensing to turn live telecom fibres into sensors: by emitting light pulses and analysing the backscatter, tiny disturbances along the fibre caused by earthquakes, ocean waves, traffic, or other phenomena can be detected.

This approach leverages already installed fibre, covering areas where dedicated sensor networks would be expensive. The research network can link scientists with cable owners, landing stations, data, and compute resources.

But fibre does not automatically become an observatory. Instrument access, fibre permissions, calibration, time synchronisation, storage, processing, and scientific validation all require dedicated protocols; sensing experiments must not degrade production communication reliability, and the data may expose sensitive activities, requiring governance.

RedCLARA is a member of a larger consortium; it does not own all the fibre and is not the sole inventor of the method. The main funding phase ends after a final dissemination event in 2026, so SUBMERSE should be described as a demonstration and a direction for collaboration, not as a permanently open regional service.

If subsequent phases can secure access rights, standards, funding, and a user community, fibre sensing could give RedCLARA a unique scientific role in geophysics, oceanography, and infrastructure resilience. At that point, the network’s value would come not only from carrying data but from measuring the world itself.

This Is a Partner Ecosystem, Not a Set of Subsidiaries

RedCLARA’s relationship map includes national NRENs, the European Commission, GÉANT, Internet2, CANARIE, development banks, EllaLink, operators, SCALAC, LA Referencia, RUTE-ALC, LNet/LACChain, universities, and thematic communities. These relationships are not legally and operationally identical.

NRENs can be members, operational partners, or both; GÉANT and Internet2 are peer networks; the European Commission is a funder and project counterparty; EllaLink is commercial infrastructure carrying capacity rights; SCALAC, LA Referencia, and RUTE-ALC have their own governance; universities may be users, node hosts, beneficiaries, or knowledge contributors.

Calling every relationship “RedCLARA” inflates central control and blurs responsibility. The organisation’s real value is precisely in coordinating assets and institutions it does not own. Responsible commentary must distinguish membership, funding, capacity, interoperability, implementation, collaboration, and project participation.

An ecosystem structure avoids a single owner but can also diffuse accountability. When a service fails, users need to know whether the problem lies with the institution, the NREN, RedCLARA, an operator, a platform, or an international partner. Clear boundaries are the only way a federation becomes a source of resilience rather than a vacuum of responsibility.

Finances Rest on Members, Projects, and Capacity Rights

RedCLARA has no verified shareholders, market valuation, or distributable profits. The statutes allow membership fees and other resources, and the organisation also obtains resources through consultancy and services, EU and development bank projects, international cooperation, partner contributions, co‑investment, capacity exchanges, and in‑kind inputs.

The most visible numbers are project budgets, not institutional revenue. ALICE was €12.5 million, ALICE2 €18 million; BELLA actions vary in scope; BELLA II is planned at €28 million, with €13 million from the EC and at least €15 million to be mobilised. Adding these multi‑year, multi‑partner amounts and presenting them as RedCLARA’s revenue is incorrect.

A significant portion of value is non‑cash, such as spectrum rights, IRUs, national fibre, colocation space, equipment, and personnel. Their real value depends on duration, maintenance, upgrade rights, and contractual limits, not on the notional project amount.

The latest full annual report describes activities, projects, and the team in detail but does not provide a complete, current audited profit and loss statement, balance sheet, reserves, membership fee concentration, or supplier exposure, nor does it precisely disclose the current staff count. This limits external analysis of financial resilience.

The core risk is a time mismatch: projects last only a few years, but backbones, identity, and platforms must run over the long term. Long-term capacity rights can preserve project value, but power, equipment, support, and staff still require recurring funding. Sustainability must come from membership fees, cost recovery, national inputs, and partner arrangements, not from treating one‑off project funds as permanent income.

The Last Kilometre Belongs to Countries and Institutions—and Is Highly Uneven

A continental backbone can be technically advanced yet still produce uneven benefits. Countries differ markedly in domestic fibre, NREN funding, government support, operator competition, cable landings, university capacity, and technical workforce; within the same country, central universities may have high‑speed connections while remote institutions still rely on fragile links.

RedCLARA can aggregate demand, negotiate capacity, train teams, and demonstrate the value of national research networks, but it cannot substitute for national investment, resolve every regulatory barrier, or operate every access line. The last kilometre still falls to institutions, NRENs, and national providers.

A “connected country” does not therefore mean every university gets equivalent service. “More than 1,300 institutions” indicates coverage but does not explain speed, availability, activeness, or use of regional services. More mature metrics would distinguish registered, active, differently‑sized connections, and actual service users.

BELLA II can expand coverage, but it also creates new obligations in places with weaker operational capacity. Sustainable inclusion requires local engineering, budgets, support, security, and identity integration. An international high‑speed link without a strong NREN can become merely a high‑capacity island.

Documentation Debt Masks Operational Maturity

RedCLARA’s public materials are rich in institutional and project history but contain conflicting metrics: seven members, eleven connected countries, and old thirteen‑country references; 10G language still sits alongside 100G‑class network descriptions; institution counts, traffic, and availability lack a consistent methodology.

The most likely reason is pages from different eras living side by side online. This does not negate the modern network but requires readers to constantly check sources, dates, and definitions. For regional infrastructure, stale documentation can become an operational risk, not just an editorial problem.

Topology maps should be dated and show contract capacity versus available capacity, protection types, operators, production status, and rights types; traffic should specify period, direction, and aggregation; availability should have scope and formula; membership information should distinguish governance, connectivity, and project participation.

More consistent disclosure would help members, funders, and users assess resilience, and it would avoid mistaking route announcements or MOUs for a live network. Transparency does not require revealing security secrets—only that what is claimed is clearly defined.

RedCLARA’s Core Achievement Is Federated Collaboration at Infrastructure Scale

RedCLARA’s most important achievement is not a particular cable, router, or project, but the creation of a long‑term regional mechanism that allows autonomous national networks to aggregate demand, run interconnections, obtain long‑term capacity rights, join global projects, and build shared services—all while retaining national governance.

ALICE turned the concept into a production network; ALICE2 shifted coordination to Latin America and sought sustainability; BELLA changed network geography with a direct South Atlantic route and long‑term spectrum; BELLA II attempts to widen coverage and transform transport into an ecosystem of identity, open science, computing, security, and experimentation.

Federation also defines its limits: RedCLARA does not own every asset, does not control every NREN, cannot guarantee every last‑kilometre, and does not have unlimited funds. Its strength comes from coordination: strong when incentives among members, funders, peers, and suppliers align; weak when responsibilities are fuzzy or national capacity is thin.

Future value depends less on the next speed record than on the institutional quality around the capacity: a more representative membership structure, long‑term contracts, transparent finances, measurable security, clear service life‑cycles, and pathways from pilot to production. The backbone is the foundation; the federal institution itself is the strategic asset.

Indicators for Whether BELLA II Becomes Long‑term Infrastructure

The first indicator is actual production. MOUs, dialogues, and procurement requirements must translate into named suppliers, accepted routes, production traffic, fault processes, and public dates. Capacity should distinguish initial minimums, line rates, protected aggregates, and available bandwidth. A 15‑year target only makes sense if operations, maintenance, and upgrade rights are also sustainable.

The second is membership structure. Today’s seven NREN members map to eleven connected countries. It will matter whether target countries become full members, long‑term partners, or institutionalised entities, and how different statuses affect votes, fees, service rights, and technical obligations. A larger network with the same recurring governance base increases operational load and keeps control concentrated.

The third is operational evidence: dated topology and traffic reports, SLA methodologies, incident disclosures, routing security policies, RPKI status, and measured diversity between EllaLink, Miami, and alternative paths. A major cable cut or route leak will test whether the logical rings have genuine physical and policy resilience.

The fourth is service transition. BIO+IA, HPC, MiLab, identity, eduLACSeg, and open science should be measured by repeat users, production allocations, support commitments, data governance, and post‑project funding, not just attendance figures. A pilot becomes infrastructure when users can depend on it and plan around it.

Finally, finance. Publishing current audited accounts, reserves, membership fee concentration, supplier exposure, and the ratio of free to restricted funds would greatly enhance credibility. The critical question is whether RedCLARA can sustain capacity and services between large European funding cycles without leaving weaker NRENs as perpetual short‑term project beneficiaries.

Evidence‑Supported Scenarios

In a successful expansion scenario, Peruvian and Central American NRENs gain long‑term, scalable links, and the regional governance base widens alongside; the combination of direct European, North American gateway, and Pacific‑side access can support larger scientific instruments and shared computing.

In a service federation scenario, connectivity fades into the background because identity, security, data, HPC, and testbeds become everyday products for users. Value deepens, but software operations, privacy, and support obligations also increase markedly.

In a governance concentration scenario, legal members remain seven while projects and connections keep expanding. The organisation can operationally cover the region but remains dependent on a small core for representation, cost‑bearing, and succession.

In a project cycle scenario, innovation is repeatedly started by grants but struggles to survive after projects end. The backbone keeps running, but higher‑layer services fluctuate with funding. This is the most realistic downside scenario without a stable cost model.

Operator co‑investment can improve sustainability through capacity exchanges, joint builds, and long‑term rights but also deepens reliance on a limited set of commercial suppliers, demanding strict governance of maintenance, exit, and upgrades.

In the “network as instrument” scenario, the SUBMERSE approach extends to more regional fibre, creating environmental and geophysical data services. It remains a forward‑looking proposition but could give RedCLARA a unique scientific role that commercial capacity alone cannot provide.

Who Controls the System, Who Bears the Risk

Control is distributed across multiple layers. Member NRENs govern the association through the Assembly and Board; the Executive Directorate and technical groups control regional implementation; national networks control domestic infrastructure and institutional access; peers like GÉANT control their own networks; funders like the European Commission influence scope through project conditions; operators, cable owners, colocation providers, and equipment vendors hold critical physical dependencies; universities and research communities create demand but do not necessarily hold direct regional governance rights.

This structure avoids a single owner but also separates decision rights from operational consequences. The Board can approve a plan where NRENs bear last‑kilometre costs; a funder designs four‑year objectives, but members inherit decades of maintenance; a single operator may hold the only realistic route into a country; users depend on a service without knowing who can fix it. The strategic layer should treat these interfaces as the primary control surfaces.

Incentives should reward long‑term contribution, not just project participation. Membership fees, fibre contributions, staff secondments, service operations, data hosting, and security response are all distinct forms of value. Countries receiving expansion support need to move towards national responsibility and recurring obligations; stronger members also need to confirm that regional solidarity does not become unlimited cost‑shifting without governance accountability.

Supplier contracts should preserve route diversity, upgrade rights, audit rights, exit, and continuity upon change of control. Long‑term spectrum and fibre rights reduce price risk but can be stranded by failures at landing stations, optical equipment, or maintenance relationships. The most critical decision is not how much capacity to buy but which dependencies become hard to reverse.

Decisions That Cannot Wait for the Next Grant

First, RedCLARA must decide whether it is primarily a connectivity federation or a long‑term operator of shared digital platforms. MiLab, identity, security, and AI testbeds create value but also introduce data, software, personnel, and liability. Services chosen for long‑term retention need clear owners, full lifecycle funding, incident response, and decommissioning rules.

Second, the mismatch between legal membership and operational coverage should be resolved. Connected countries need clear status, voice, and a pathway to membership or other accountable arrangements. Informality during expansion may be convenient, but it leaves long‑term ambiguity over who governs, who pays, and who can exit.

Third, financial transparency should become a governance tool. Audited accounts, reserves, membership fees, restricted funds, and supplier concentration determine whether the network can withstand a funding interruption or a cable incident. Without this evidence, members cannot responsibly take on long‑term obligations.

Fourth, RedCLARA should publish a continuity architecture covering route diversity, incident escalation, configuration and registry backups, identity and platform recovery, failover routes for cable cuts, and succession arrangements for project services. When a grant, supplier, path, or national institution fails, the regional network should still sustain essential functions.

The irreversible risk is not a single project delay but the federation accumulating critical services and long‑term dependencies faster than it builds recurring authority, funding, and operational evidence. The irreversible opportunity is the opposite: using BELLA rights, NREN governance, and global partnerships to build a regional infrastructure commons that no single country could create alone. Leadership should preserve the distributed character while writing responsibilities clearly enough that “shared ownership” does not mean “shared ambiguity.”

The Long‑term Significance of a Shared Regional Backbone

RedCLARA demonstrates an infrastructure autonomy built on cooperation, not isolation. Latin American research gains greater agency, not because the organisation owns every fibre, but because national networks jointly control a regional layer, hold strategic long‑term capacity, and participate in global research partnerships on more equal terms. Commercial infrastructure dependencies remain, but mission, routing relationships, and upgrade decisions are no longer entirely left to the general internet market.

This model has meaning beyond research. It shows that institutions with fewer or uneven resources can aggregate demand without surrendering national autonomy, and that grants can be converted into rights and organisations that outlast project cycles. It also reveals the limits of federation: coordination cannot infinitely compensate for weak domestic networks, opaque finances, or unclear accountability.

The decisive choice is whether the next phase merely expands the map or matures the institution. More countries and faster links matter, but the enduring assets are the governance, operations, identity, data, security, and research collaboration systems built around them. If that system is transparent, financially sound, and genuinely regional, RedCLARA will remain one of Latin America’s most important shared digital infrastructures; if not, the backbone can still carry traffic while higher‑layer ambitions remain dependent on the next external funding scheme.