Summary

  • Ecuadortelecom S.A. remains the registered holder of AS27738, but current public fixed-service documents identify CONECEL S.A. as the contracting and operating company behind Claro.
  • AS27738 is visibly active and sizeable, yet current routing observations show one adjacent autonomous system; that is a logical dependency signal, not proof that every customer circuit shares one physical route.
  • Claro publishes substantial international fixed-network capacity and broad maintenance footprints, but it does not publish the customer-level duct, aggregation, power and restoration evidence needed to certify truly independent dual entry.

At dawn, two entrances become one question

At dawn outside a commercial building in Guayaquil, an honest resilience survey would begin with a deliberately mundane exercise. One technician would stand by a fibre label at the front service entrance. Another would follow the second label around the loading side. They would not stop when the cables disappeared into separate sleeves. They would keep tracing both paths to the first street chamber, pole line or handhole where the routes could be inspected against construction drawings. If the two tails met under the same concrete lid twenty metres from the wall, the building would have two entrances but only one civil failure domain.

That scene is hypothetical. No public document reviewed for this article maps an Ecuadortelecom or Claro enterprise circuit at that level, and it would be misleading to present a particular chamber, duct or building as observed. The point is methodological: physical independence is a chain of facts, not a product label. It must survive the route from customer equipment through risers, entrance facilities, street ducts, feeder cables, aggregation nodes, metro transport, upstream interconnection, electrical supply and the people authorised to repair each segment.

The distinction matters because the public identity of the network is already layered. Ecuador’s regional internet registry still assigns AS27738 to Ecuadortelecom S.A. in its current registration record. Claro’s contemporary legal pages, by contrast, place current fixed offers under CONECEL S.A. while retaining a separate historical Ecuadortelecom document collection. A customer can therefore encounter the Claro brand, sign a CONECEL document, receive addresses originated by AS27738 and still lack a route diagram showing which legal or operating party controls the fibre beneath the pavement.

The public offer is substantial. Claro advertises dedicated corporate internet as permanent, symmetric and backed by a 1:1 committed information rate, with fibre or radio access and 24-hour support. It markets MPLS data services for private local and international links and SD-WAN for centralised policy and access optimisation. Those are commercially meaningful capabilities. None of the public product pages, however, identifies a pair of customer-specific street routes, names two independent aggregation facilities or warrants that two access circuits will avoid a common splice chamber.

That gap is not evidence that the network lacks diversity. It is evidence that diversity cannot be inferred from the materials available to a public reader. There may be private route packs, engineering drawings and contractual commitments supplied during a corporate tender. There may also be sites where fibre and radio genuinely terminate in different metro and power domains. The unknown is site-specific, and a responsible assessment must keep it unknown until the operator or an independent surveyor produces the relevant artefacts.

The first question for an enterprise buyer is therefore not, “Can Claro deliver two links?” It almost certainly can in many locations. The better question is, “At what exact point do the two links become independent, and at what exact point do they converge again?” Everything else in the resilience discussion follows from that answer.

The Ecuadortelecom name survives in routing, not on today’s fixed-service contract

Ecuadortelecom has a documented regulatory history in Ecuador. ARCOTEL maintains a page for the company’s 2015 request to renew its concession for telecommunications services. The company name also remains attached to AS27738 at LACNIC. Those records establish that Ecuadortelecom was a licensed telecommunications entity and that its name continues to identify a live internet-routing resource. They do not, by themselves, establish the company that signs a new enterprise access contract in 2026, employs the field crew, owns a particular cable or receives the monthly payment.

Current customer-facing documents point elsewhere. Claro’s fixed-telephone service page identifies the service and its guide with CONECEL, while the company’s enterprise personal-data notice names CONECEL as the entity associated with services offered through the corporate portal. A current carrier-service annex is a CONECEL form that records the medium, speed, origin, destination, installation fee, monthly fee and sharing level for a link. A current internet-service annex likewise names CONECEL and sets out residential, small-business and corporate service fields.

The safest description is therefore narrow. Ecuadortelecom S.A. is the registered AS27738 holder and a historically licensed fixed-telecommunications company. CONECEL S.A. is the company named on the current public fixed-service and enterprise contractual material examined here. Claro is the public brand under which those services are marketed. Public materials do not clearly explain whether Ecuadortelecom remains a separately active contracting company, whether its operational functions have been fully integrated into CONECEL, or which entity holds title to every physical asset associated with AS27738.

Historical documents show that the relationship was operationally close long before the current contracts. An ARCOTEL decision about a March 2015 Ecuadortelecom fixed-telephony interruption says the outage followed a fibre cut in a transport network supplied by CONECEL. A 2016 ARCOTEL resolution records an infrastructure-sharing agreement between CONECEL and Ecuadortelecom. A mirror of a filed CONECEL financial statement describes a long-term fibre usufruct, service and maintenance arrangement entered into with Ecuadortelecom in December 2014. The mirror is useful corroboration, but it is not a substitute for a current title register or an executed customer contract.

Those records support a bounded inference: the two companies’ fixed-network operations have shared transport, infrastructure and commercial arrangements. They do not prove that every present-day Ecuadortelecom-originated route depends on one CONECEL cable, nor do they prove that Ecuadortelecom owns none of the plant. The legal ownership of individual ducts, fibres, shelters, routers and power systems remains undisclosed in the public materials reviewed.

The distinction is more than corporate housekeeping. If an enterprise buys a primary and backup circuit under one Claro relationship, the party responsible for repairing the street fibre may differ from the party managing the IP service, the building access agreement or the upstream capacity. A service-level commitment is only as useful as the named party’s control over the failed component. Buyers should ask the contracting entity to identify the asset owner, operator and restoration owner for each path segment, rather than treating the brand, legal entity and autonomous-system name as interchangeable.

ARCOTEL’s January 2026 fixed-internet market bulletin reinforces the present-day naming boundary: its December 2025 provider chart attributes 11.9 per cent of fixed-internet accounts to CONECEL, not to Ecuadortelecom. That statistic is a market-share measure, not a network map, but it is consistent with the contractual evidence. In current public reporting, CONECEL is the operating name attached to the retail fixed footprint; Ecuadortelecom remains most plainly visible in the routing layer and historical record.

AS27738 is active, large and logically single-adjacent

AS27738 is not a dormant label. LACNIC’s registration identifies Ecuadortelecom S.A. in Guayaquil and lists a fixed-network technical contact using a Claro corporate address. RIPEstat’s announced-prefix observation returned hundreds of IPv4 and IPv6 prefixes during its July 2026 observation window. The exact count changes with the query window and global routing state, so it should not be treated as a permanent inventory. It does show that the autonomous system is actively originating a broad address portfolio.

The striking feature is adjacency. RIPEstat’s AS-neighbour observation showed one adjacent autonomous system, AS23487, identified as CONECEL. BGP.tools also described AS27738 as active with AS23487 as its upstream at the time of review. These independent routing views support a strong logical conclusion: from the public global routing perspective, AS27738’s external reachability is currently exposed through CONECEL’s autonomous system rather than through a visibly diverse set of external AS neighbours.

That conclusion has limits. One BGP neighbour does not mean one router, one fibre, one city, one cable landing or one international carrier. A pair of autonomous systems can interconnect across many ports, chassis, buildings and geographically separated routes while presenting one logical adjacency to the internet. Conversely, multiple BGP neighbours can ride the same conduit or facility. Routing diversity and physical diversity are related but non-equivalent properties.

Public routing-security data adds another layer without resolving the physical question. RIPEstat reports a valid route-origin authorisation for AS27738’s 190.130.128.0/17 block. That is useful evidence that the observed origin is authorised under the Resource Public Key Infrastructure. It reduces one class of routing-origin risk. It says nothing about whether two enterprise tails share a feeder cable, whether an aggregation router has dual power or whether a flood can isolate both paths.

Self-reported peering data illustrates the same boundary. Ecuadortelecom’s PeeringDB network record names Claro Ecuador, reports an estimated traffic level of 300–500Gbps and describes a selective policy. It does not list public exchange or facility entries on the viewed record. The traffic band is an operator-supplied estimate, not an audited capacity figure, and the absence of listed facilities is not evidence that the network has no facilities or exchange presence. It simply prevents a public reader from using that record to locate independent interconnection sites.

Third-party scale estimates should be read just as carefully. APNIC Labs estimated roughly 1.09 million users behind AS27738 on 29 June 2026, or about 9.61 per cent of the country’s measured user population. The Internet Society’s Ecuador internet-resilience profile also places Ecuadortelecom among the country’s significant networks. These are measurement-based estimates, not audited subscriber counts. They indicate that failure at the AS or a major shared aggregation layer could have material reach, but they do not identify which customers, products or cities sit behind any particular prefix.

The economic implication of the single visible AS adjacency is still important. If AS27738 obtains external reachability through AS23487, a customer purchasing two AS27738-delivered circuits cannot assume upstream organisational independence merely because the access media differ. Fibre plus radio may protect against a trench cut but still converge on the same CONECEL routing domain. Two fibres may enter from opposite streets yet terminate on the same provider edge router. An SD-WAN overlay may shift traffic between links but cannot manufacture an upstream that was never provisioned.

None of this makes AS27738 fragile by definition. CONECEL may operate geographically redundant interconnection, multiple international carriers and resilient internal transport beneath the single public AS relationship. Indeed, its capacity disclosures show multiple external links. The precise unknown is the mapping between AS27738, AS23487, customer access nodes and those international gateways. Until that mapping is disclosed for a customer design, the routing evidence supports a dependency question, not a verdict of physical single-homing.

What the enterprise offer promises—and what it leaves blank

Claro’s corporate internet proposition is clearer about service quality than about route geometry. The published product promises dedicated, permanent and symmetric access with a 1:1 committed rate, delivered over fibre or radio and monitored with round-the-clock support. That is materially different from a contended consumer service. A committed information rate gives the customer a contractual performance reference, and the choice of media creates options for access design.

The carrier annex captures several useful commercial attributes. It asks for the origin and destination, medium, speed, installation price, recurring charge and sharing level. Those fields can define what is being bought. They do not require the sales document to disclose the cable route, street chamber sequence, aggregation device, provider-edge site, upstream AS, electrical feed or restoration depot. A filled annex could therefore be complete as a commercial order and still be incomplete as a resilience certificate.

The internet annex makes the product boundary more visible. It distinguishes service categories, records upload and download rates, and states a 2:1 sharing level for the listed offers. A separate 2026 filing for fixed internet plans up to 850Mbps describes GPON small-business plans and a wired last mile. These plan rates are retail product speeds. They are not evidence of dedicated enterprise port capacity, PON split ratios at a given optical line terminal, backhaul headroom or performance after another link fails.

MPLS and SD-WAN add useful service layers. MPLS can separate customer traffic and connect sites across an operator’s footprint. SD-WAN can monitor application conditions and steer traffic among available underlays. Neither product name proves that the underlays avoid common civil infrastructure. If both circuits enter the same metro ring through the same feeder duct, an overlay can react only after both have disappeared.

The public service descriptions also leave ownership boundaries open. A fibre circuit may include an operator-owned tail, leased duct, shared pole, third-party building riser and customer-powered router. A radio backup may depend on a rooftop lease, one local electrical panel and a remote base site fed by the same metro aggregation hub as the fibre. The contracting carrier may control only some of those assets directly. The rest may be governed by access agreements and repair queues that matter greatly during a regional outage.

Service area is another field that should not be overread. Claro markets nationwide enterprise connectivity and supplies customer-premises equipment across Ecuador. Its public distributor and network description refers to hundreds of thousands of fixed and wireless services and a large radio and fibre footprint. Those marketing figures support national scale, but they do not establish that every city offers two independent enterprise routes or that the same redundancy standard applies at every building.

A buyer seeking high availability therefore needs a design schedule alongside the commercial annex. That schedule should identify route A and route B by entrance, duct, aggregation point, provider edge, transport segment and upstream handoff. It should say whether each element is owned, leased or shared; whether both services are restored by the same crew; and whether maintenance on one element can place the other at risk. The operator can redact security-sensitive coordinates while still providing enough geometry and naming to support an engineering review.

The absent details should not be treated as a defect unique to Claro. Telecom product pages rarely publish customer-specific civil plans. The analytical error would be to convert a normal absence of public detail into an assumption of independence. A 1:1 rate, an MPLS label, an SD-WAN appliance and two access media answer different questions. None answers the dawn survey’s central question: where do the paths first meet?

Dual entry is not route diversity

“Dual entry” describes the building edge. “Route diversity” describes the network beyond it. The two can coincide, but one does not guarantee the other. A resilient design must remain disjoint through a sequence of failure domains, and the sequence should be explicit enough that a customer can test it.

Inside a building, the two services should avoid the same riser, communications room, patch frame and power strip. At the perimeter, they should use different sleeves or poles and should not enter the same unprotected chamber immediately outside. In the street, they should avoid the same duct bank, bridge crossing, river crossing and construction corridor. Farther out, they should reach different aggregation devices or at least different line cards and chassis, with a documented point at which controlled convergence becomes acceptable.

The distinction between topology and ownership is equally important. Two cables owned by different companies can share a municipal conduit. Two services sold by one operator can use genuinely separate civil routes. A map with differently coloured lines proves little unless it identifies the physical structures beneath those lines and the party that can access them. “East route” and “west route” are descriptions; chamber identifiers, pole sequences, facility names and controlled redactions are evidence.

Ecuador’s historical record supplies a concrete warning about assuming independence from service labels. In the 2015 fixed-telephony case, the regulator recorded that Ecuadortelecom service was interrupted after a fibre cut in transport supplied by CONECEL. That event is eleven years old and cannot establish the current architecture. It does demonstrate the failure mechanism: one company’s customer service can depend on another company’s transport, and the dependency may become visible only when a physical cable is cut.

Infrastructure sharing creates both efficiency and concentration. The 2016 regulatory record of sharing between CONECEL and Ecuadortelecom, and the 2014 financial-statement description of fibre use and maintenance, show why a customer must ask about the asset layer rather than infer separation from corporate names. Shared fibre arrangements can expand coverage and lower capital costs. They can also cause two nominally different services to enter the same maintenance domain.

Access technology is not a shortcut. Fibre and radio are diverse against some hazards but not all. A radio path can avoid a trench cut yet share the customer’s power, rooftop access, metro router and network-operations process with the fibre path. GPON and dedicated Ethernet can use different access electronics but share a feeder duct or hub. Two mobile links can use distinct radio cells while both depend on the same backhaul or commercial power substation.

The burden of proof changes with the customer’s consequence of failure. A small office may reasonably accept best-effort diversity and rapid mobile failover. A hospital, bank, emergency service or industrial plant may need route drawings, power autonomy, tested failover and named restoration owners. The word “redundant” is too broad to serve all of those cases.

There is also a time dimension. Two paths that were diverse when installed can converge after roadworks, emergency repairs, fibre grooming or network consolidation. A carrier may reroute one tail into a shared duct without changing the customer-facing service name. True assurance therefore requires change control: the operator should notify the customer when a physical or logical alteration reduces the agreed separation.

No reviewed public document provides that route-level evidence for a specific Ecuadortelecom or Claro enterprise customer. No public map identifies two independent customer entrances, ducts, aggregation sites and upstreams end to end. This remains the central unknown. It should be resolved by customer-specific engineering evidence, not by extrapolating from nationwide coverage or aggregate capacity.

Claro’s hub notices expose aggregation failure domains

Maintenance notices are unusually revealing because they show how operational work is grouped. Claro’s subscriber information page publishes planned interventions for fixed services. The notices do not reveal complete topology, but they identify named hubs, access technologies, maintenance windows and geographic areas that can experience simultaneous impact.

A notice for 9 July 2026 describes software and patch work on the NEX8A PE-AGG hub in Machala. It names CONECEL as the provider and warns of potential effects on fixed telephone, internet and television services over HFC and GPON during a six-hour overnight window, with up to four hours of impact. This is direct evidence that one aggregation intervention can touch multiple access technologies and services in one city. It is not evidence that every Machala enterprise circuit traverses that router.

A 30 June notice covers work associated with the Collaloma, Ponceano and Samanes hubs. It anticipates a potential two-hour interruption to GPON services within a six-hour window and names areas across Quito and other Andean cities, as well as Guayaquil’s Tarqui parish. The wide list does not mean one physical router serves every named location; the work package may encompass several devices or coordinated changes. It does show that the operator manages failure and maintenance domains that can span numerous communities.

The pattern appears again in the Villaflora and Durán hub notice for 23 June, which lists affected GPON areas across Pichincha and cities including Durán, Quevedo, Milagro, Portoviejo, Babahoyo, Cuenca, Azogues, La Libertad and Loja. A 19 June Almendros, Huancavilca and Inca notice similarly connects planned work with GPON impact in parts of Guayaquil and Quito.

These notices matter for enterprise design in three ways. First, they demonstrate that a named hub or coordinated hub work package can be a shared operational boundary across large service areas. Second, they show that access technology alone does not isolate customers from aggregation maintenance. Third, they identify CONECEL as the operating party issuing notices and managing the intervention, reinforcing the distinction between the current operator and the Ecuadortelecom AS name.

They do not answer several crucial questions. The notices do not publish the number of affected customers, the precise fibre routes, the protected status of each enterprise circuit, the failover outcome for dual-homed sites or the reason some areas are grouped together. They do not state whether a four-hour potential impact represents expected downtime, a worst-case allowance or a period during which redundancy is reduced. They also do not identify the field teams, spare equipment or escalation path if the planned work causes an unplanned fault.

For a customer, the practical test is whether route A and route B appear in the same maintenance domain. If both terminate on the same PE-AGG hub, two diverse entrances may still lose external reachability during a router failure or software event. If they terminate on different hubs but both hubs are included in one coordinated intervention, the customer needs to know whether work is staged to preserve one live path. A maintenance calendar is therefore part of resilience, not merely an administrative notice.

The notices are strong evidence of real operational scope but only medium-strength evidence for topology. They reveal where simultaneous impact is possible without proving the exact architecture beneath it. That is enough to reject a simplistic assumption that two access tails automatically remain independent through aggregation. It is not enough to declare a specific customer design unsafe.

International capacity is substantial but not a customer resilience certificate

CONECEL publishes a monthly fixed-network international gateway report under Ecuador’s quality-reporting framework. The June 2026 report lists seventeen international access links for each day of the month. Fourteen are shown at 100,000Mbps, one at 200,000Mbps and two at 10,000Mbps, for a reported aggregate contracted capacity of 1,620,000Mbps, or 1.62Tbps.

The links carry labels associated with multiple international providers, including Cogent, GTT, Sparkle and Telia, and with transport contexts such as AMX, Telxius and PCCS. That is meaningful evidence that CONECEL’s fixed international gateway portfolio is not represented by one named carrier circuit. The daily traffic ratios also show that aggregate utilisation was well below total contracted capacity during the reported month.

The report does not, however, publish an N-1 or N-2 survival model. One individual link reached a daily ratio above 90 per cent during the month, while aggregate utilisation varied substantially. A simple subtraction of average traffic from total capacity would overstate resilience because traffic cannot necessarily move freely to every remaining link. Ports may terminate in different cities or facilities, routes may have policy constraints, and traffic engineering may reserve headroom unevenly.

The capacity is also scoped to CONECEL’s fixed international gateways, not specifically to AS27738. The report does not map each link to AS27738 or AS23487, identify which prefixes can use which gateway, or state how much capacity is committed to residential, small-business, enterprise or wholesale traffic. It therefore cannot be converted into a per-customer bandwidth guarantee.

PeeringDB’s self-reported 300–500Gbps traffic band for AS27738 should not be added to the 1.62Tbps figure. One is an estimated network traffic range attached to an autonomous-system profile; the other is a regulatory table of contracted international access capacity for CONECEL’s fixed network. They use different scopes and definitions. Treating them as additive would create a number that no source supports.

Retail access rates require the same discipline. An 850Mbps GPON plan is a customer product speed, not evidence that 850Mbps remains available through a gateway failure. A 1:1 corporate committed rate is a service commitment at the contracted interface, not proof of international path separation. Capacity can be installed but unpowered, lit but unavailable to a particular route, operational but reserved for another service, or usable only under normal conditions.

The strongest defensible conclusion is that CONECEL reports a large, multi-link international fixed-network portfolio with significant observed aggregate headroom in June 2026. The unknowns are the geographic distribution of those links, their common landing and metro dependencies, failover policy, protected capacity, reserved headroom, AS27738 eligibility and customer-level availability under simultaneous failure.

That distinction matters in procurement. A customer asking whether the operator has “enough capacity” should specify the failure condition: enough after loss of the largest link, after loss of one gateway facility, after a cable cut affecting several links, or after loss of commercial power at an aggregation site. The answer should state which capacity is sold, which is reserved, which is physically available and which has been tested under the named condition. Aggregate terabits are impressive; resilience lives in the subtraction case.

Facilities and power remain separate proof obligations

Facilities are often used as shorthand for resilience, but the presence of a data centre does not prove that a customer access circuit traverses it. Claro markets housing services with controlled environment, connectivity and security features. Current housing terms identify a CONECEL data centre on the Durán–Boliche corridor in Guayas. Those documents establish a real current facility and a service counterparty. They do not establish that AS27738’s upstream interconnection, a given enterprise customer’s provider edge or either of two access routes is located there.

A third-party facility page for Cirion’s Quito 2 data centre lists Claro among providers and CONECEL among peering parties. It also publishes an installed power figure for that facility. This is useful evidence of commercial or interconnection presence at a named Quito site, but the wording is the facility operator’s representation. It does not identify an AS27738 port, the capacity of any CONECEL connection or a customer route through the building.

The safest facility map is therefore sparse: a current CONECEL housing location near Durán and a third-party report of CONECEL/Claro presence at a Quito data centre. Public material reviewed for this article does not locate the NEX8A PE-AGG equipment, the Collaloma, Ponceano, Samanes, Villaflora, Durán, Almendros, Huancavilca or Inca hub equipment at street-address precision. Hub names indicate operational geography, not verified facility coordinates.

Power evidence is broader but similarly non-specific. ARCOTEL’s 2024 accountability report records CONECEL’s statement that critical infrastructure was supported with generators and battery banks during Ecuador’s electricity crisis, together with spending on generators, batteries, power plants, spares and fuel. That supports the existence of a serious backup-power programme. It does not publish the site inventory, autonomy hours, battery condition, refuelling priority, generator loading, dual-utility arrangements or the definition of “critical.”

An enterprise path crosses several power domains. The customer router and optical terminal may depend on one building panel. A basement communications room may have a generator while the external radio unit does not. An access cabinet may use batteries, an aggregation hub may have a generator, and a data centre may have a separate protected system. If both circuits depend on one local panel or one unprotected street cabinet, resilient core facilities cannot prevent the outage.

Ecuador’s National Emergency Telecommunication Plan sets out a general framework for coordination and restoration during disasters. It is not an operator-specific assurance and does not supply customer route or autonomy data. Its relevance is institutional: recovery during a widespread emergency depends on coordination, access and prioritisation as well as equipment redundancy.

The public record therefore supports a measured inference. CONECEL operates or uses significant facilities and has invested in backup power for infrastructure it considers critical. It does not support a claim that every Ecuadortelecom/AS27738 enterprise access path has dual power, a specified autonomy period or a geographically separate termination. Facility resilience and route resilience must be proved separately, then tested together.

For a dual-entry design, the buyer should request a power-domain schedule that starts at the customer equipment and follows each path to the first protected core node. It should identify utility feeds, battery autonomy, generator coverage, refuelling assumptions and any shared panel. A route diagram without power annotations is incomplete; a Tier-designed data-centre brochure without a route diagram is equally incomplete.

Restoration depends on crews, access rights and named ownership

Resilience is usually sold as architecture and experienced as labour. When a cable is cut, the practical outcome depends on who detects the fault, who can enter the chamber, who has the splice plan, where replacement cable is stored, whether police or road authority access is required, and which customer receives priority. None of those questions can be answered by a BGP graph.

Claro’s institutional description says the company has more than 3,400 direct workers and a nationwide service presence. Its corporate internet offer promises 24-hour support. Those statements indicate organisational scale and an around-the-clock customer contact model. They do not disclose how many fixed-network field technicians are stationed in each province, whether civil repair is outsourced, where splicing crews are based, or what mobilisation time applies to an enterprise fault.

The planned-maintenance notices show that CONECEL can coordinate work across named hubs and broad geographic areas. Planned work, however, is the easiest restoration condition: engineers have a defined window, a known task and pre-positioned staff. An unplanned cut after a landslide, vehicle collision, fire, theft or power failure presents different constraints. Access may be unsafe, a third-party duct owner may need to attend, and both nominally diverse links may compete for the same crew.

The 2015 Ecuadortelecom outage decision is revealing because it identifies a transport dependency on CONECEL. It does not provide enough detail to reconstruct the repair, and its age prevents direct comparison with present practice. It nevertheless shows why restoration ownership must be explicit. The customer-facing service and the damaged transport can sit under different contractual relationships, while the customer experiences one outage.

For each path segment, a high-assurance design should name four roles: asset owner, network operator, maintenance contractor and customer escalation owner. One organisation may fill all four roles, but that should be stated rather than assumed. If a municipal conduit, utility pole, building landlord or wholesale carrier controls access, the service-level clock should explain how those dependencies are handled.

Recovery targets also need a starting point. Mean time to repair can begin at alarm detection, ticket creation, technician dispatch, site access or physical work. A four-hour target measured after site access can produce a much longer customer outage if permission takes six hours. The public materials reviewed do not state route-specific repair targets for Ecuadortelecom/AS27738 enterprise circuits, nor do they publish historical performance against such targets.

The same uncertainty applies to spares. A router can be protected with an on-site module, while a buried cable may require a length, closure, jointing crew and civil excavation. Generator resilience depends on fuel delivery during a national power emergency. Radio resilience can depend on tower access and replacement antennas. A resilience schedule should distinguish equipment restoration, cable restoration, power restoration and logical rerouting because each has a different owner and clock.

Local support labour is therefore part of network economics. Geographic coverage creates revenue opportunity, but it also creates a dispersed maintenance burden. The hub notices span coastal and Andean cities, implying an operating footprint that must be supported across distance and terrain. Public documents do not reveal whether staffing and spare placement scale evenly with that footprint. An enterprise buyer outside Quito or Guayaquil should ask for local mobilisation evidence, not rely on a national headcount.

The most useful recovery commitment is a tested procedure. The operator and customer should simulate loss of each entrance, each access device and the shared aggregation point identified in the design. The test should record detection time, failover behaviour, ticket ownership, communications cadence and restoration path. Without that exercise, a diagram describes intended architecture; it does not prove operational recovery.

The customers at risk are concentrated at convergence points

The impact of a network failure is not evenly distributed. It concentrates where many services converge: a building riser, feeder cable, optical line terminal, provider-edge router, metro ring, gateway facility, upstream interconnection or power system. The maintenance notices make that concentration visible at the hub level, while AS27738’s public routing relationship makes it visible at the autonomous-system level.

ARCOTEL’s market data attributes a material share of Ecuador’s fixed-internet accounts to CONECEL. APNIC’s separate measurement estimates a large user population behind AS27738. The figures use different methods and scopes and must not be combined. Together, they establish only that both the current CONECEL fixed business and the Ecuadortelecom routing identity are significant enough for shared failures to matter beyond one enterprise.

Residential users may experience a hub intervention as lost internet, television or fixed voice. Enterprises can face wider consequences: payment terminals stop, cloud applications become unreachable, remote sites lose corporate connectivity, call centres disconnect and security monitoring degrades. A customer with two links may still fail if both depend on the same domain-name resolution, firewall, power panel or provider edge. Network diversity is necessary but not sufficient for application continuity.

The concentration can also be commercial. If the same operator supplies both links, manages both customer routers and owns the escalation path, recovery is simpler to coordinate but less independent organisationally. Buying a second carrier can diversify operations, yet two carriers may lease the same local fibre or meet in the same facility. Neither single-provider nor multi-provider design is automatically superior; the physical and operational facts decide.

Service area needs precise language. The current maintenance notices demonstrate that CONECEL operates GPON and other fixed services in numerous named cities and parishes. They do not prove that Ecuadortelecom S.A. is the legal owner of those access assets, that all traffic uses AS27738, or that every area supports the enterprise products described on the national website. Public national availability should not be converted into building-level serviceability.

Customer impact also depends on failure timing and duration. Planned maintenance during an overnight window may be tolerable for an office and unacceptable for a factory operating continuously. A short routing reconvergence may interrupt real-time sessions even if average availability remains high. A long civil repair can exhaust batteries that easily cover a brief power transfer. Availability percentages conceal these mechanisms unless paired with event definitions.

The strongest risk signal in the evidence is not a known single point of failure. It is the number of convergence points that remain undisclosed. The public record shows one visible external AS neighbour, named aggregation work packages, shared historical infrastructure, a multi-link international portfolio and broad backup-power investment. It does not connect those elements into a customer-level dependency graph.

That missing graph is where procurement should focus. A bank branch in Cuenca, a clinic in Machala and a warehouse in Durán may buy the same branded product but face different ducts, hubs, facilities, power conditions and crew travel times. Resilience cannot be awarded at company level and inherited by every site. It must be assessed per location, per path and per failure condition.

The evidence an enterprise buyer should demand

The first deliverable should be a route-diversity statement for each site. It should show both building entrances, internal risers, street chambers or poles, duct corridors, feeder cables, aggregation nodes, provider-edge sites and upstream handoffs. Exact coordinates can be restricted for security, but the document must still identify shared segments and the distance over which the paths remain physically separated.

The second should be an ownership and restoration matrix. For every segment, the contracting party should name the asset owner, operator, maintenance contractor and escalation owner. It should disclose leased or shared infrastructure and state whether the two paths rely on the same civil contractor, spare stock, access permit or field crew. This is the practical answer to the Ecuadortelecom–CONECEL identity split.

The third should be an aggregation and upstream schedule. It should state whether the circuits terminate on different ports, cards, chassis and sites; whether both enter AS27738; where AS27738 connects to AS23487; and whether failure of one interconnection or one CONECEL facility leaves both services reachable. The public single-adjacency observation makes this question necessary, but it cannot answer it.

The fourth should be a capacity survival table. Normal contracted rates should be separated from capacity available after loss of the largest link, one gateway site or one metro route. The operator should identify reserved headroom and any traffic-engineering constraint that prevents all surviving capacity from serving the customer. CONECEL’s 1.62Tbps June portfolio is useful context, but the buyer needs the portion and path that remain usable in the agreed failure case.

The fifth should be a power-domain schedule. It should follow each link from customer equipment to the first protected core node and record utility feeds, battery autonomy, generator coverage, fuel assumptions and shared panels. It should distinguish design autonomy from tested autonomy. National investment in generators and batteries is not a substitute for the site schedule.

The sixth should be a maintenance and change-control commitment. It should prevent simultaneous planned work on supposedly independent paths, or require an agreed temporary backup when such work is unavoidable. It should require notice if route grooming, civil relocation or equipment consolidation creates a new shared failure domain. Diversity can decay silently unless someone is responsible for preserving it.

The seventh should be a recovery playbook with measured clocks. It should define when detection, dispatch, access, repair and customer communication timers start. It should identify local crew and spare locations without overstating sensitive detail. It should state how restoration is prioritised when a regional event affects many customers and whether a wholesale dependency can pause the service-level clock.

The eighth should be a witnessed failover test. The customer should disconnect each access path in turn, test failure of customer equipment where possible and verify that critical applications continue over the surviving route. A desk review cannot reveal every policy, addressing, firewall or domain-name dependency. Results should be recorded and repeated after material network changes.

The evidence does not justify saying that Ecuadortelecom or CONECEL lacks resilient enterprise infrastructure. AS27738 is active, CONECEL reports multiple international links, Claro markets dedicated access and the operator demonstrates a broad fixed-service footprint. Nor does the evidence justify certifying two-path independence. No public material closes the chain from separate building entrances through separate civil works, aggregation, power and upstream restoration.

That is the durable finding. The corporate names, routing scale and aggregate terabits are real, but they sit at different layers. Enterprise continuity depends on the joins between those layers. Until the joins are documented, a second circuit is an additional service—not yet proof of an independent route.