Summary

  • VOLZ said in September 2024 that 14 large communications nodes in Kyiv had batteries capable of supporting power for 120 hours without grid electricity, with generators also available. The page does not publish node-level load tests, maintenance records or an end-to-end SLA.
  • The operator also describes more than 3,000 kilometres of fibre in Kyiv and the surrounding region, routes over three Kyiv bridges, AS12963, UA-IX participation and international connectivity. These are separate layers, not substitutes for one another.
  • A decision-grade continuity claim needs a ledger that joins each critical customer site to its serving node, tested runtime, physical path, aggregation and upstream dependencies, customer power, incident owner and contractual remedy.

Five days is a component runtime, not an outcome

On 25 September 2024, VOLZ wrote that it had 14 large communications nodes in Kyiv districts. It said powerful uninterruptible-power batteries had enough capacity to sustain power for 120 hours without electricity, and that generators were available in emergencies. Five days is materially more informative than an undefined claim of “backup power”. It gives a buyer something to test.

The public statement does not say which load was used to calculate the runtime, when each battery was installed, how degradation is measured, whether all 14 nodes have identical systems, or which generators serve which locations. It does not disclose fuel autonomy, refuelling arrangements, field staffing or the last node-by-node failover result. Nor does it state that 120 hours is a contractual availability target.

That boundary does not make the disclosure weak. It makes the next request obvious. A customer should ask for the serving node, equipment covered by the backup system, tested load and runtime, test date, alert path and transition plan from battery to generator. A rated capacity is an input. A dated test under representative load is operational evidence.

The remaining path matters even if the node stays powered. Customer premises may lose electricity. A building entrance, duct, optical amplifier, aggregation site or remote end may fail. A generator may operate while a shared fibre is damaged. The useful proposition is therefore not “120 hours of Internet”. It is “a named component was designed and tested to remain powered for a stated period, while the other components have their own controls”.

VOLZ exposes several layers that must be joined

VOLZ says it has more than 3,000 kilometres of its own fibre in Kyiv and the surrounding region. Its network page names routes over the South, North and Metro bridges. Those facts describe geographic reach and possible routing options. They do not establish that two purchased services avoid common ducts, buildings, power feeds, aggregation routers or operating teams.

The company history adds logical-network detail. VOLZ says it built an L2 ring using ITU-T G.8032 during 2017-2019 and a new L3 core with connections up to 200 Gbit/s during 2020-2022. A separate post says the Kyiv-Lviv channel was doubled to 200G in November 2023. Its corporate-internet page also speaks of direct channels from Frankfurt and an international-channel capacity of 50 Gbit/s.

These figures should not be combined into one capacity total. They describe different parts of the system, at different dates and with incomplete scope. A core interface rate, a city-to-city channel and an international-channel figure answer different questions. None reveals present utilisation, spare capacity, route preference or failover performance.

For procurement, the better map has a physical column and a logical column. The physical column records the two fibre paths, bridge or duct exposure, entrances, nodes and power. The logical column records aggregation, origin ASN, upstreams, exchange sessions, route preference and DDoS controls. A service is meaningfully diverse only when both columns avoid the same relevant failure domain.

AS12963 is observable, but visibility is not availability

RIPE NCC lists the legal-name company as a Ukrainian member. RIPE RDAP records AS12963, named VOLZ, as active and links it to the same legal entity. UA-IX lists VOLZ Scientific-Industrial Firm Ltd. as AS12963 with two addresses at its named Kyiv VEGA site. These records establish a credible identity and network-resource chain.

RIPEstat supplies a dated observation. At query time 28 August 2026 08:00 UTC, it saw AS12963 at all 327 IPv4 and all 320 IPv6 RIS peers in the returned set. It reported 48 originated IPv4 prefixes covering 33,792 addresses, one originated IPv6 prefix covering 65,536 /48 units, and 196 observed neighbours. The companion response returned 49 announced-prefix records, including aggregates and more-specifics.

That is evidence of a widely visible control-plane footprint at the captured moment. It is not evidence that a customer circuit, battery, generator or international path was available. RIPEstat itself notes that routes with very low visibility are excluded. Prefix and neighbour counts also do not equal customers, links, facilities, traffic, capacity or physical diversity.

The UA-IX record has a similar boundary. It identifies an exchange participant and addresses; it does not prove that a particular session was established at capture, how much traffic it carried or whether the path was independent of another route. A public route may remain visible while one access area is dark. Conversely, a local service may work while a remote destination fails.

Security controls form another failure domain

VOLZ says its DDoS service uses Arbor Sightline and a Threat Mitigation System. The public page describes FlowSpec, BGP blackhole, traffic scrubbing and a three-second activation claim for the stated layer-3/4 detection scenario. These are meaningful control descriptions because they identify who can observe traffic and initiate mitigation.

They are not an incident record. The page does not publish scrubbing capacity, attack volumes, false positives, customer-impact minutes or the circumstances in which traffic is blackholed rather than cleaned. A business that depends on availability should distinguish power continuity from congestion, routing error and defensive traffic discard. Each can produce the same user symptom through a different mechanism.

The operator’s incentives also change across these layers. Adding batteries protects equipment. Adding route options can reduce dependence on one interconnect. DDoS mitigation preserves scarce link and service capacity. None of the three automatically funds a field repair, customer-site power or an application’s recovery. The contract and incident process must preserve these distinctions.

The missing product is an evidence ledger

A compact continuity ledger would begin with the customer’s critical site and the service ordered. It would name the last-mile route, serving node, aggregation site, origin ASN, exchange and upstream dependencies, customer equipment and remote service. Each component would have an owner, failure mode, backup control, last test, observed result and escalation contact.

For the 120-hour claim, the ledger should state whether the figure is design capacity or tested runtime; the supported load; battery age and condition; generator hand-off; fuel and field access; and the service scope that depends on that node. For fibre diversity, it should identify common ducts, bridges, entrances and aggregation. For routing, it should record intended and observed sessions without treating a registry object as a live test.

VOLZ’s public material is useful because it exposes enough of the stack to ask precise questions. The due-diligence mistake would be to flatten those disclosures into one adjective such as “resilient”. A buyer does not need fewer claims. It needs each claim tied to its failure domain, evidence owner and remedy.