AP Voglu Internet Service makes a local connectivity bill depend on upstream routes and field repair

Summary

• Voglu Internet Service presents a compact fibre-broadband business in Old Dhaka. Its current coverage page names nine neighbourhoods, its office is in Laxmibazar, and its current residential offers run from 25 Mbps for Tk700 a month to 70 Mbps for Tk2,000, plus an undefined "unlimited" tier.
• Current operating evidence is credible. AS139033 originates 512 IPv4 addresses, its routes were visible to all 327 IPv4 peers in a recent RIPE RIS observation, a customer portal answers from Voglu's own address space, and a June 2026 Dhaka probe reached an address inside the network.
• The visible route is narrow. Independent routing views show only one adjacent network, Alpha Broadway System's AS64074. That provider has several upstreams and exchange connections of its own, but their diversity does not prove that Voglu receives physically independent handoffs or automatic failover.
• The access network remains largely undocumented. Voglu says 90% of its customers use optical fibre and that it serves more than 1,000 customers, yet it publishes no fibre-route map, pole or duct rights, aggregation sites, power reserve, busy-hour capacity, repair targets or tested recovery design.
• Regulatory status needs clarification rather than assumption. A BTRC list dated 18 December 2024 identifies Voglu as a Sutrapur Upazila/Thana licensee but prints a validity date of 29 January 2024 and a next-renewal date of 30 January 2024. A current ISP association entry and live operations do not substitute for a newer regulator confirmation.

A local network with a very specific urban footprint

Voglu Internet Service does not describe a national network. Its coverage page names nine parts of Old Dhaka: Laxmibazar, Sutrapur, Banglabazar, Islampur, Shankhari Bazar, Gendaria, Narinda, Koltabazar and Banianagar. Its contact page places the office at 63 Easel Heaven Tower in Laxmibazar, Dhaka 1100. The company's home page calls the business an independent provider for homes, businesses and corporate users in Old Dhaka and claims more than 1,000 customers.

That is a commercially coherent footprint. The named areas sit close enough together for a local operator to aggregate customers through a relatively compact access plant and dispatch technicians from one office. The address also corresponds closely to the regulatory and industry records. The Internet Service Providers Association of Bangladesh member list records Voglu under membership A-472, gives a Laxmibazar address and describes the licence class as Upazila/Thana. A Bangladesh Telecommunication Regulatory Commission list dated 18 December 2024 places the licensee in Sutrapur at Easel Heaven Tower.

The footprint is useful precisely because it is not vague. It establishes a neighbourhood provider whose physical challenge is urban distribution rather than long-distance rural coverage. A connection in Shankhari Bazar or Narinda must get from a customer's room to a building entry point, through risers and street distribution, toward an optical aggregation point and onward to the routing edge. Every turn creates an ownership and repair boundary.

The company can sell the service, configure customer equipment and operate its network, but it may not own the pole, building riser, underground route, electrical feed or long-haul transmission carrying the traffic away from Old Dhaka.

The coverage list is not a route map. It does not reveal whether each neighbourhood is served by a continuous fibre ring, separate spurs, leased capacity, partner distribution, or a mixture. It does not show which streets are orderable, whether every building can be reached, or how many customers share an aggregation port. A named neighbourhood is evidence of an intended commercial area, not proof of service at every address within it. The right scale for assessing Voglu is therefore nine advertised neighbourhoods and one named office, with the exact street-level plant still undisclosed.

Voglu's about page calls it the largest broadband provider in Old Dhaka and says 90% of customers have moved to optical fibre. Both are first-party claims. They suggest a substantial migration away from older access media, but no public customer count by technology, third-party market share or dated completion report accompanies them. The statements are best read as operating signals: the company wants current and prospective users to understand it as a fibre provider, and it says a remaining tenth of customers still require an upgrade. They cannot establish market leadership or the precise amount of fibre in service.

That remaining tenth matters. If the claim is accurate, the subscriber base is not technically uniform. Customers on different access media may experience different fault patterns, available speeds and restoration methods. A fibre cut requires optical testing and splicing; an Ethernet fault may require replacing copper cable or a switch port; a wireless segment would introduce line-of-sight and radio-power dependencies. Voglu does not say what the non-fibre connections use. A recovery plan therefore cannot assume that one tool kit, one spare or one technician profile restores every subscriber.

What the monthly plan actually promises

The company's current package page publishes six offers. The first five advertise up to 25, 30, 40, 50 and 70 Mbps for Tk700, Tk800, Tk1,000, Tk1,400 and Tk2,000 a month. The sixth costs Tk4,000 and is labelled "unlimited" without a numeric line speed. Every tier advertises an optical-fibre connection and access to torrent, file-transfer, gaming, live television, movie and streaming servers.

These offers are a window into regional ISP economics. They combine a global-internet rate with local or on-net content that may travel a much shorter path. The value perceived by a household can therefore exceed what the international transit allotment alone would suggest. A cached film, local game server or domestic file source can be fast even while an overseas cloud service is constrained. Conversely, a plan can feel disappointing during international congestion despite excellent throughput to a nearby server. One headline number cannot describe both paths.

The words "up to" transfer part of the capacity question from the package to the network. An access port may be capable of 70 Mbps while the shared aggregation or upstream link cannot deliver that rate to every customer simultaneously. The relevant engineering measure is not the sum of all sold headline speeds, because residential customers do not peak together at full rate, but neither is it the port speed alone. The operator must forecast concurrency, video demand, local-cache use, international traffic, protocol overhead and growth, then hold enough headroom for faults and maintenance.

Voglu does not publish a contention ratio, monthly data allowance, fair-use rule, guaranteed minimum, international-versus-domestic split or service-level objective on the package page. It also does not define what the Tk4,000 "unlimited" plan makes unlimited: volume, local speed, international speed or something else. That absence does not mean such terms are unavailable at sale, but a reader cannot calculate the usable capacity from the public offer.

The price context is moving quickly. The regulator's earlier One Country, One Rate page described maximum shared-broadband prices of Tk500 for 5 Mbps, Tk800 for 10 Mbps and Tk1,200 for 20 Mbps, and said the minimum speed would become 10 Mbps from December 2021. By January 2026, state-owned BTCL published GPON offers including 20 Mbps for Tk399, 25 Mbps for Tk500, 50 Mbps for Tk800 and 100 Mbps for Tk1,050. Those are not direct like-for-like comparisons: availability, installation, support, local content, route quality and service conditions differ. They do show why a small provider must compete on more than the number printed beside a monthly fee.

Voglu's lowest listed plan, 25 Mbps for Tk700, costs more than BTCL's published 25 Mbps plan. Its 30 Mbps plan at Tk800 offers less headline speed than BTCL's 50 Mbps plan at that price. A local operator can still win if it reaches a building the larger provider does not, installs quickly, maintains local content, answers calls and repairs a severed drop promptly. Those advantages are labour- and network-intensive. They also disappear quickly when a technician is unavailable or a shared upstream congests.

The commercial page therefore poses the core question. Does Voglu merely have equipment capable of delivering each plan under light demand, or can it maintain the advertised experience at busy hour and after a component fails? No public utilisation chart answers it. The correct assessment is that the plans are orderable signals of a current business, while the capacity behind them remains undisclosed.

The licence defines the geographic and infrastructure boundary

Bangladesh's ISP licensing guideline separates nationwide, divisional, district and Upazila/Thana licences. It says an Upazila/Thana licensee is authorised to provide service within the administrative area of the particular Upazila or Thana. That framework aligns with the Sutrapur entry in the regulator's list and with Voglu's compact Old Dhaka marketing.

The same guideline makes the ownership boundary unusually important. It says an ISP should lease or sublease transmission from licensed nationwide telecommunication transmission network operators, with alternatives governed by infrastructure-sharing rules where such service is unavailable. It limits last-mile connectivity to approximately three kilometres in metropolitan areas and requires compliance with local-authority directions. It also calls for connection to a licensed International Internet Gateway for internet bandwidth and to a National Internet Exchange for domestic inter-operator traffic.

In other words, being the customer's ISP does not imply ownership of the entire path. Voglu may own optical line terminals, switches, customer optical units, patch panels, cables or other equipment, but the public evidence does not show which assets are owned and which are leased. The transmission beyond its local point of presence may depend on another licensed carrier. Domestic exchange access may be direct or reached through another network. International traffic must ultimately traverse the country's gateway structure. Each contract and physical handoff can add redundancy or concentrate risk.

The guideline also requires service quality in accordance with regulator directions and says access networks should remain compatible with next-generation and IPv6 standards. It does not turn every commercial statement into a guaranteed speed, but it places quality and technical evolution inside the licensed responsibility. Voglu's IPv6 allocation, discussed below, shows preparation at the address-resource level; current public routing does not show a normal production IPv6 origin.

There is an unresolved date in the public licence record. The December 2024 BTRC list identifies licence number 14.32.0000.702.46.708.19.369, dated 27 December 2021, but gives a validity date of 29 January 2024 and a next-renewal date of 30 January 2024. Those dates had passed before the list itself was compiled. The current ISP association page still displays the same licence number, and Voglu's site and network are active in 2026, but neither fact is a replacement for a current BTRC document.

It would be wrong to infer illegality from a dated list alone. The record could reflect a renewal pending at the compilation date, a data-quality issue, a subsequent renewal not found in the public material, or another regulatory state. It would be equally wrong to describe the 2021 licence as unquestionably current. The accurate conclusion is narrower: Voglu had a named Sutrapur Upazila/Thana licence in the regulator's December 2024 list, but present validity requires newer confirmation.

That confirmation matters operationally. A customer may care first about whether the line works, but suppliers, landlords, transmission carriers and public authorities need a clear legal counterparty. Licence continuity can affect contracts, permissions and investment. For a network whose advantage depends on fast local action, uncertainty at the authorisation boundary is itself worth resolving.

From the customer socket to the routing edge

A fibre-broadband session begins inside a building, not in a routing registry. Customer equipment needs electricity. An optical drop needs a protected path from the premises to a distribution point. Splitters, connectors and splices must stay clean and within optical-loss limits. A feeder then reaches an optical line terminal or other aggregation equipment, which needs power, cooling, secure space and a working uplink. Switching and subscriber-management systems must authenticate the user before traffic reaches an internet edge.

The company supplies enough clues to identify this chain but not to draw it. It explicitly markets optical-fibre connections and says most subscribers use fibre. It provides separate telephone numbers for complaints, new lines and general queries on its contact page, and it links to a customer portal. On 10 July 2026, that portal answered at 103.138.182.13, an address inside Voglu's own registered IPv4 space. The main marketing site resolved to a different Bangladeshi hosting network, which is a sensible separation: a public information site can remain reachable even if Voglu's access network has a local fault.

The portal response is stronger operating evidence than an old directory entry. It shows a customer-facing service still reachable on the company's network. It does not disclose how many users logged in, what functions were active, whether the portal shares a site with the routing edge, or whether it has redundant power. It also cannot prove that every advertised neighbourhood has working access. It is one live component in the operating chain.

At the network boundary, the APNIC registration for AS139033 identifies the name VIS-AS-AP, country code Bangladesh and description Voglu Internet Service. It was registered on 30 April 2019. APNIC separately records 103.138.182.0 through 103.138.183.255 as a portable allocation associated with Voglu, a total of 512 IPv4 addresses. These resources allow the operator to present an independently numbered edge rather than relying entirely on addresses lent by an upstream.

The RIPEstat routing-status view observed three IPv4 routes covering those 512 addresses as of 10 July 2026. It reported that all 327 IPv4 route-collector peers in that snapshot could see the network. The three routes are the covering 103.138.182.0/23 and the two more-specific /24s. They do not represent three independent pools of addresses; the /24s sit inside the /23. Nor do three announcements establish three physical exits. They are routing entities describing overlapping reachability.

The bgp.tools view of AS139033 also lists the /23 and its two component /24s, classifies the network as an access or "eyeball" network and marks it active. It shows valid route-origin authorisation for the IPv4 announcements. RIPEstat's RPKI validation result likewise says the covering route is valid for origin AS139033.

That validation is useful but limited. It means the cryptographic authorisation for the route matches the observed origin, reducing one class of accidental or malicious route-origin error. It does not prove that the fibre is intact, capacity is sufficient, the route is geographically optimal, or the customer-facing service is licensed and supported. RPKI protects an origin statement; it does not provide end-to-end resilience.

The strongest recent reachability detail comes from IPinfo's page for 103.138.183.0/24. It records a traceroute from a Dhaka probe on 18 June 2026 that entered AS139033 and reached 103.138.183.20 with low local latency. It also reports a responding address in that /24. A separate AS139033 overview recorded responsive IPv4 and IPv6 addresses and describes the network as a small ISP.

A single probe is not an availability study. It samples one destination at one time from one city. The low latency is consistent with a local Dhaka path and current equipment, but it says nothing about evening congestion or a customer's optical signal. It nevertheless joins the portal, current route visibility and active website as independent signs that Voglu is operating a reachable network in 2026.

One visible neighbour is the central route risk

The same public routing evidence that establishes operation exposes a narrow dependency. The RIPEstat neighbour view shows one observed neighbour for AS139033: AS64074. bgp.tools and IPinfo also identify Alpha Broadway System as the only visible adjacent or upstream network. Public collectors can miss private or backup sessions, but three independent views agree on the path they can see.

One adjacent network is not necessarily one cable. Voglu could buy two circuits from Alpha Broadway System, take delivery at different sites, use diverse fibres, or hold a standby service that is invisible until activated. Alpha Broadway could also carry Voglu over multiple upstream routes. None of those arrangements is published. From outside the network, every visible path converges first on the same autonomous system.

Alpha Broadway's own public topology is broader. Its bgp.tools page shows several upstream relationships and lists connections at BDIX, Summit NIX and ISPAB-NIX. The BDIX member page confirms Alpha Broadway as a entity at Bangladesh Internet Exchange, while the Internet Society IXP tracker lists a 10 Gbps port and three peering locations for AS64074.

Those facts improve the likely onward options after traffic enters Alpha Broadway. They do not confer direct exchange membership on Voglu. Voglu's own PeeringDB network record lists no public exchange point, no interconnection facility, no disclosed prefix count and no multiple-location requirement. Its open peering policy is a stated policy, not proof of a session. The record was last materially updated in July 2022, making its blank fields informative about disclosure but not definitive about current physical connectivity.

This distinction is essential. A supplier with three upstreams can protect its core and still deliver a customer over one lateral fibre into one building. Two upstream names on an invoice can share a duct, pole line, power room or router. Conversely, one supplier can provide genuinely diverse handoffs if the circuits enter separate sites and terminate on independent equipment. Autonomous-system adjacency reveals commercial and routing concentration; only route and facility documentation can reveal physical independence.

For Voglu, the first resilience test is therefore not simply "buy a second upstream." It is to establish two usable paths from the local network edge whose vulnerable portions do not coincide. That can mean separate fibre routes, separate handoff buildings, separate powered edge devices and, ideally, distinct provider networks. The failover must preserve customer addressing, authentication and enough capacity for busy-hour demand. A backup that carries only management traffic or collapses under peak load is installed redundancy but not usable redundancy.

The current topology also creates a support dependency. When customers cannot reach overseas services, Voglu must determine whether the fault lies in its access plant, its edge, the handoff to Alpha Broadway, Alpha Broadway's onward network, a gateway, a submarine or terrestrial international route, or the destination. With only one publicly visible neighbour, there is no observed alternate route around a provider fault. Clear escalation contacts and measurements from both sides of the handoff become as important as spare fibre.

IPv6 is allocated, but normal production use is not established

APNIC allocated 2404:d240::/32 to Voglu in May 2019. That is an enormous logical address space compared with the 512-address IPv4 allocation, as is normal for IPv6. Address count should not be confused with customer capacity: IPv6 is designed to allocate large subnet blocks, and the existence of a /32 says nothing about traffic volume or equipment throughput.

Current routing views do not show a stable production IPv6 announcement. bgp.tools lists zero originated IPv6 prefixes. RIPEstat recorded a small number of observations around the end of June and beginning of July 2026, including the /32 and many /48s, but its 10 July routing-status snapshot counted zero announced IPv6 /48 equivalents and only one of 321 IPv6 collector peers seeing the network. That combination suggests transient, experimental or very low-visibility activity rather than broadly available IPv6 service.

IPinfo reported that an IPv6 address in Voglu's allocation responded from a Chennai probe, which is another useful signal that some part of the space has been configured. It does not establish native dual-stack access for residential customers. Public confirmation would require stable global announcements, measurements from multiple collectors, customer-side IPv6 addressing and successful end-to-end tests over time.

The gap matters for both capacity and resilience. IPv6 can reduce dependence on scarce public IPv4 addresses and simplify some forms of end-to-end reachability. It also adds another routing plane, firewall policy, customer-equipment requirement and support burden. A network that advertises IPv6 intermittently without monitoring it can create hard-to-diagnose failures where applications prefer a weak IPv6 path over a healthy IPv4 one. The responsible conclusion is that Voglu holds suitable resources and has shown limited activity, while public evidence of mature customer IPv6 service remains absent.

Installed fibre is not the same as usable capacity

Voglu's claim that 90% of customers use optical fibre describes access medium, not service capacity. Fibre can carry enormous bandwidth, but the usable rate is set by the optical equipment, splitter design, uplink ports, aggregation, subscriber policy, local caches and external transit. A thin uplink can constrain hundreds of excellent optical drops. A saturated splitter or switch can constrain one building while the core has spare capacity.

The number of IPv4 addresses is similarly unhelpful as a capacity gauge. Five hundred and twelve public addresses could support fewer than 512 directly addressed devices, many thousands of translated customers, infrastructure interfaces, servers or a mixture. Cloudflare Radar's AS139033 overview has recently estimated several thousand users, while Voglu says more than 1,000 customers. The figures measure different things and use methods that are not public at customer level. Neither reveals paid lines or simultaneous demand.

The right capacity questions are physical and temporal. How many optical line terminal ports are installed? At what split ratio? What is the uplink from each aggregation point? How much domestic and international bandwidth is committed? How high does utilisation run between 8pm and 11pm? What happens when one uplink or chassis is removed? How many new lines can be installed before another feeder or port is required? Voglu publishes none of these figures.

Local media services can materially reduce external load. The package page links to local addresses and Alpha-branded media domains, consistent with content being available near the access network. Keeping popular traffic local can improve video start times and lower paid transit use. It can also make a speed test or local download look much better than access to a distant service. Customers need both: local efficiency for common content and sufficient external capacity for work, cloud storage, software updates, education and communication.

The 2024 outage of one of Bangladesh's submarine systems showed the distinction at national scale. The Internet Society's account found that terrestrial links and local caches helped Bangladesh continue operating, though some internationally hosted services slowed. A local ISP can remain "online" while the quality of particular paths deteriorates. Capacity reserved for failure conditions determines whether rerouting feels like resilience or merely a softer outage.

Voglu's public offers therefore demonstrate installed commercial capability up to at least a 70 Mbps residential tier. They do not demonstrate that 70 Mbps remains available internationally at busy hour, that every fibre customer has the same optical design, or that failover capacity equals normal capacity. Usable capacity should receive credit only after measurement.

Failure path one: a fibre route is cut or ordered away

Old Dhaka's density gives a local operator short distances but difficult rights and working conditions. Fibre may cross crowded poles, building fronts, narrow roads, ducts, risers and shared utility corridors. A single feeder can pass many customers before branching. Road work, utility excavation, fire, construction, unauthorised cutting, vehicle contact or a landlord dispute can disconnect it.

Dhaka has a documented history of tension around overhead communications cable. A Prothom Alo editorial on cable removal described tangled internet and television cables on utility poles, efforts to sever them, and disagreement over responsibility for moving services underground through NTTN operators. In March 2024, Dhaka metro authorities ordered internet and television cables removed from above the MRT-6 alignment. Those examples are not Voglu incidents, and the metro corridor is not evidence about Voglu's route. They show that local access cables in Dhaka can be affected by public works and safety orders even when the electronics remain healthy.

The regulator's metropolitan last-mile limit and local-authority requirement make route management part of the service. A provider needs an inventory showing where each feeder runs, which poles or buildings it uses, who can authorise access and where an alternate can be placed. If a municipal or utility project requires removal, advance notice is useful only when the operator has materials, permissions and labour to move the circuit.

A ring is the familiar answer, but the word alone proves little. Two fibres that occupy opposite logical directions can share the same pole for most of their length. A ring can also converge at one powered switch or one building entrance. When that shared segment is cut, both directions fail. Physical route records should identify common crossings, bridges, ducts, poles, entrances and aggregation rooms. A failover test should then break one direction under realistic load and verify customer sessions, not just a green link light.

Repair speed depends on fault location. An optical alarm at the central equipment may indicate loss, but technicians still need to identify which span failed. An optical time-domain reflectometer can estimate distance; field crews need route records to translate that distance into a place. They may need pole access, traffic control, landlord entry, a ladder, a lift or a splicing enclosure. A replacement cable must be safely installed, spliced, tested and protected. Every undocumented turn adds minutes or hours.

Voglu advertises a complaint telephone number and 24-hour support, but it does not publish a mean repair time, escalation target, crew count or spare-cable stock. The compact service area should enable rapid dispatch from Laxmibazar. It also means one local event can affect a large share of the footprint if feeders converge near the office or edge site. Short distance is an advantage only when route knowledge and access rights are current.

Failure path two: power disappears at one layer of the chain

Fibre is passive between powered endpoints, but a fibre service is not. The customer's optical unit and Wi-Fi router require electricity. Building switches, optical line terminals, aggregation routers, authentication systems, portal servers and the upstream handoff require electricity. A battery at the edge cannot keep a dark customer router online, and a household battery cannot compensate for an unpowered aggregation site.

Bangladesh's telecom sector continues to experience this dependency. In July 2025, a BTRC disaster-planning meeting discussed power outages, network failures, damaged fibre and microwave links, fuel shortages and technician access. In April 2026, Banglalink offered compensation to customers affected by network disruption during electricity outages. These are national and mobile-sector signals, not evidence that Voglu suffered the same failures. They identify practical hazards that any powered access network in Dhaka must plan around.

Backup power must be evaluated by layer and duration. A small uninterruptible power supply may bridge a brief interruption at the office. Longer outages require larger batteries, a generator, fuel, ventilation, maintenance and safe transfer. Distributed building equipment may have less reserve than the main edge. If technicians cannot reach a site or a landlord locks the equipment room, a fully fuelled generator elsewhere is irrelevant.

No public Voglu page identifies batteries, generator capacity, runtime, fuel arrangements or power diversity. The office address may or may not contain the routing edge. The customer portal's presence inside Voglu's address space may or may not share power with access equipment. The upstream handoff may or may not terminate in the same building. Inferring a single-site design would be speculation, but so would granting credit for diverse power.

The proper test is a timed inventory. Each powered node needs its normal draw, battery condition, expected runtime, alarm path and restoration priority. Generator-backed sites need load tests and fuel plans. Customer communications should explain that an operator-side network can remain available while an unpowered home cannot use it, and should specify compatible backup options for the optical unit and router. None of this requires a large provider; it requires disciplined local operations.

Power also interacts with congestion. When one site fails, surviving equipment may inherit customers or traffic. Its electrical load and uplink utilisation rise at the same time. A backup path sized for idle monitoring may stay electrically alive but deliver unusable performance. Recovery design must therefore test power and traffic together.

Failure path three: the one observed upstream is lost

If the Voglu-to-Alpha Broadway adjacency disappears and there is no hidden alternate, customers can retain local optical synchronisation while losing the wider internet. The portal or local media may still respond if they sit inside the reachable local network. That partial availability can complicate diagnosis: some services work, others do not, and the access light stays green.

The event can originate in several places. Voglu's edge router may fail. The optical handoff may lose signal. A fibre between the two networks may be cut. Alpha Broadway may withdraw the route, lose power at the delivery site or suffer a configuration error. Voglu's valid RPKI authorisation reduces the risk of an invalid origin being rejected, but it cannot maintain a failed session.

Alpha Broadway's multiple onward connections are valuable only after Voglu traffic reaches it. If the first handoff is the failed component, diversity deeper in the supplier network does not help. The decisive controls are a second physical handoff, a second edge device, and preferably a distinct provider relationship. A cold backup that requires manual readdressing can take too long; a configured session should be tested by withdrawing the primary and watching routes converge.

Routing failover also needs capacity. The primary and secondary routes may have different latency and international paths. If the backup is smaller, traffic management must preserve essential service without causing repeated session resets. The company should know how long the route takes to appear globally, whether customer addresses remain valid and whether domestic traffic still reaches an exchange efficiently.

The public record provides no evidence of that second path. PeeringDB lists no Voglu exchange presence or facility, and route collectors show one neighbour. The company may hold private arrangements that are not visible, but resilience cannot be awarded on possibility. The present external assessment is single-neighbour exposure with unknown circuit count and unknown physical path diversity.

Local support labour is part of the product

Voglu's website emphasises skilled technical staff and 24-hour telephone support. For a compact ISP, that claim is economically important. The person who answers a local complaint may know the building, the splitter and the pole. A national call centre can have more staff but less street-level knowledge. Local labour can turn geographic concentration into faster diagnosis.

The same concentration can create key-person risk. Optical access, routing, power and customer equipment require different abilities. A field installer may replace a drop but not diagnose BGP. A network engineer may restore a route but not be authorised to climb or work near electrical infrastructure. A fibre splicer needs specific equipment and clean consumables. After a widespread event, multiple calls can arrive while the same small team is repairing the shared cause.

The Bangladesh Digital Ecosystem Country Assessment reported that providers faced physical-security problems including damaged towers and cut fibre, and described the cost and disorder associated with repeated cable repair. That is sector-level evidence, not a Voglu loss figure. It explains why a low monthly tariff must finance more than bandwidth: route inspection, spares, transport, protective enclosures, tools and people who can work safely in a crowded city.

For Voglu, useful resilience evidence would include the number of on-call teams, skills available per shift, contractor agreements, escalation contacts with the transmission supplier, configured spare routers and optical modules, reels of compatible fibre, splice closures, power supplies and customer units. It would also include access arrangements for nights, holidays and public emergencies. None is public.

The customer-facing phone structure is still a positive signal. Separate numbers for complaints, new lines and general queries can keep sales traffic away from fault reports. A reachable website hosted outside Voglu's own network provides an alternate place for status information. The site does not currently expose a public status page or incident history, however. During a wide outage, telephone calls may still overwhelm the team unless a separate mobile-accessible channel gives scope and estimated restoration time.

Repair performance is where the local operator justifies its price. The ability to send someone who recognises a particular lane or building can outweigh a modest difference in Mbps. But locality is not the same as guaranteed availability. It must be backed by enough people, authority, tools and spare parts to handle two simultaneous failures.

Who loses service when one component fails

The affected population depends on where the chain breaks. A damaged customer drop may disconnect one household. A building feeder or switch may affect dozens of homes and shops. An aggregation failure can remove a neighbourhood. Loss of the edge or sole visible upstream can affect the entire Voglu footprint while leaving lights on at customer premises.

The consequences are broader than streaming. Voglu explicitly markets to homes, businesses and corporate users. A household can lose remote work, education, messaging and access to cloud files. A shop can lose payment connectivity, ordering and customer communication. A small office can lose hosted applications, voice and remote administration. The article does not identify any particular institution as a customer; the public customer list is not available. These are the functions exposed by the service classes Voglu advertises.

Partial failure can be especially confusing. Local media may continue while an overseas work platform fails. IPv4 may work while an unstable IPv6 route delays connections. One building can be dark while another street remains healthy. A speed test to a nearby server can pass while international capacity is congested. Support staff need path-specific measurements so that "the internet is slow" becomes a testable fault domain.

The company is also exposed financially. A mass outage creates refund pressure and consumes labour while revenue remains fixed by monthly plans. Emergency fibre, overtime, fuel and supplier escalation can erase the margin from many low-cost subscriptions. If customers have competing fibre in the building, a prolonged repair can become permanent churn. This is why route diversity and spares are not luxuries added after growth; they protect the economics of the installed base.

What the public evidence proves, and what it does not

Voglu's current operating case is credible. Its website is active and presents specific prices, neighbourhoods and published contact points. Its portal answers from its own IPv4 allocation. APNIC records the company-controlled network resources. Multiple routing observers see three valid IPv4 announcements covering 512 addresses through one neighbour. A June 2026 probe reached the network from Dhaka. These independent signals are difficult to reconcile with a purely dormant name.

Scale is less certain. The company says it serves more than 1,000 customers and that 90% use optical fibre. Cloudflare Radar estimates a user population in the low thousands. Those numbers are directionally compatible, but they are not the same measure and neither is independently audited in the material reviewed here. No subscriber filing, line count, revenue figure or equipment inventory is public.

Physical design is weaker still. There is no street-level fibre map, aggregation-site list, pole or duct agreement, optical split design, power reserve, edge location, handoff address, busy-hour graph or restoration record. The named coverage area and fibre marketing support a local access network; they do not prove a ring or a second route.

External redundancy is visibly thin. One observed neighbour is a material concentration. Alpha Broadway has onward diversity and exchange presence, but the evidence does not show that Voglu receives more than one delivery path or that traffic can bypass Alpha Broadway. PeeringDB's lack of Voglu facilities and exchanges reinforces the absence of public proof, though an incomplete record cannot prove absence.

Regulatory continuity is unresolved. The 2024 list's past validity date should trigger a request for a current BTRC licence or renewal notice, not a declaration that the provider lacks authority. The current ISP association membership is supportive but belongs to a different institution. A publication-ready assessment must keep those records separate.

Unofficial traffic estimates, geolocation pages and commercial topology sites are useful signals, not final authority. They can suggest live customers, Dhaka location and adjacency. They cannot prove exact subscriber counts, legal status, fibre ownership or service quality. The operating conclusion rests on convergence across company pages, APNIC registration, global route visibility, a live portal and recent reachability, not on any one estimate.

The evidence that would demonstrate recovery rather than merely promise it

The strongest improvement would be a simple current network statement. It should identify the licensed service area, current licence validity, number of aggregation sites, primary and backup upstreams, exchange path, and whether the backup arrives over a physically separate route. Exact sensitive paths need not be published; an independent engineering review could verify separation without exposing street coordinates.

For the access plant, Voglu should be able to show that major neighbourhood feeders either form a ring or have a documented replacement route. A test should open one link during busy hour, record the customers moved, measure packet loss and throughput, and confirm that optical and authentication systems recover. A paper ring earns little credit if both legs share the same pole line or entrance.

For upstream resilience, the operator should withdraw the primary handoff and demonstrate that all three IPv4 announcements remain reachable through a separately powered and routed connection. The test should record convergence time, domestic and international latency, available capacity and any sessions that fail. IPv6 should be either carried stably with the same discipline or withheld from customers until it can be monitored reliably.

For power, every critical site needs a measured runtime under actual load, not a battery label. A generator plan needs fuel, safe operation and tested transfer. Distributed building equipment should be included. Customer guidance should explain what remains powered by the provider and what the household must back up.

For field repair, the useful measures are time to detect, time to locate, time to gain access, time to splice or replace, and time to restore every affected customer. Spares should match the installed equipment. At least two people or contracted teams should cover critical skills so that illness, simultaneous faults or unsafe conditions do not stop recovery.

Finally, capacity should be stated as usable service. Busy-hour utilisation, oversubscription policy, international versus domestic performance and failover headroom would allow customers to interpret the package speeds. A local network does not need to publish commercially sensitive costs to show that a 70 Mbps plan survives the loss of one uplink.

A credible operator with an unproven safety margin

AP Voglu Internet Service is not merely a stale routing label. The evidence as of 10 July 2026 supports a working small ISP in Old Dhaka: a live commercial site, a reachable customer portal inside its own address space, globally visible and authorised IPv4 routes, a recent Dhaka traceroute, a specific office and nine named service areas. That is enough to rate current operation as credible.

The evidence does not support a strong resilience claim. The only publicly observed adjacent network is Alpha Broadway System. Voglu discloses no exchange, interconnection facility, physical route, backup power, spare capacity or repair target. Its fibre and customer figures are self-reported, IPv6 appears intermittent rather than mature, and the latest located regulator list leaves licence renewal after January 2024 unclear.

The final network evidence grade is therefore Medium. Operation is supported by several current technical and commercial signals; physical topology and recovery remain too opaque for Strong. The grade would fall if the portal or routes disappeared, or if a current regulator record contradicted operation. It would rise with current licence confirmation, two independently verified handoffs, tested route and power failover, measured busy-hour headroom and documented field-repair performance.

For a customer, the monthly bill may look simple: choose a speed, install an optical connection and call a local number when it fails. The underlying product is a chain of permissions, power, fibre, routing contracts and human response. Voglu's opportunity is that its compact footprint can make that response genuinely local. Its risk is that the same compactness can place many customers behind one cable corridor, one room or one upstream. The difference can be established only by showing where the network can bend without breaking.