FriktoriaNet George Sgouridis trading as Friktoria.com - Data Center Services. has to prove whether marketed data-centre capacity can survive power and carrier constraints

Summary

• Friktoria.com has a verifiable public routing identity. The RIPE RDAP record for AS204932 names FriktoriaNet and lists George Sgouridis trading as Friktoria.com - Data Center Services. as the registrant at Koutarelia 54, Volos, while RIPEstat's AS overview marks the AS announced on 12 July 2026.
• The visible network is small. RIPE announced-prefix data shows one IPv4 route, 185.246.164.0/24; RIPE routing status counts 256 IPv4 addresses, no visible IPv6 route and two observed neighbours; CAIDA AS Rank likewise sees one prefix, 256 addresses and two providers.
• The facility story is more detailed than the independent proof. Friktoria's Volos facility page claims a 350 square metre DC01VOL room, renewable-grid supply, diesel generation, UPS clusters, A/B rack power lines, controlled temperature and humidity, CCTV, access control, fire systems, several named internet providers, more than 400 Mbps of external links and 99.999% nominal uptime.
• Retail capacity is explicitly sold in small increments. The colocation page offers 1U, 5U, 10U, 21U, full-rack and footprint options, typically with one 10/100/1000 Mbps switch port, one IPv4 address and either 230VAC single-phase or 400/230VAC three-phase power. The dedicated-server page and VPS page show hosted-infrastructure products that appear to sit on the same operating base.
• The proper public grade is cautious. AS204932 proves an active routed edge, and the company gives unusually specific facility and power descriptions for a small host. The same record still lacks third-party facility certification, live-load disclosure, generator fuel runtime beyond the stated UPS bridge interval, cooling failover evidence, route-diverse carrier entry proof, customer failover records and evidence that marketed capacity can keep serving customers through a utility, cooling or carrier incident.

The company is visible, but the capacity question is narrower

The first useful conclusion is that Friktoria.com is not merely a stray hosting label. The network record, the company website and the product catalogue point to an operating Volos business with a small data-centre and hosted-infrastructure offer. The RDAP autnum record gives AS204932 the name FriktoriaNet, records the AS as active, lists George Sgouridis trading as Friktoria.com - Data Center Services. as registrant, and places the organisation and named contact at Koutarelia 54 in Volos. The same record shows a 2018 registration date and a 2022 last-change date. That is a stronger anchor than a hosting brand that appears only on a marketing page with no matching number-resource identity.

The second conclusion is more important for buyers. An active ASN and a public server catalogue do not prove that every advertised power, cooling and carrier claim works during a bad day. Friktoria's own pages describe a facility that is serious enough to merit scrutiny: A/B rack power lines, three-phase diesel generating sets, UPS clusters, controlled temperature and humidity, multiple named internet providers, physical security, fire systems, colocation units and hosted server products. Those are not vague words. They are operationally meaningful claims.

But most of the detail is company-origin, and the public record does not show an independent facility audit, a current customer reference, a failover test report, a generator runtime table, a carrier entrance diagram or an incident history.

That boundary defines this profile. The question is not whether Friktoria has a footprint. The question is whether the footprint can be treated as resilient data-centre capacity rather than a small server-room business with a live routed edge. For a customer using a single 1U colocation slot, one dedicated server or a low-cost VPS, the distinction may look academic until power drops, a UPS branch trips, a cooling unit fails, or one upstream carrier withdraws routes. For a business that uses the site as its only Greek hosting location, the same distinction becomes commercial risk.

The public evidence therefore supports a disciplined middle position. Friktoria has enough visible infrastructure to be researched as a real data-centre services provider, not dismissed as a name-only network. It does not have enough public operating proof to let customers treat its marketed uptime and A/B resilience as settled facts. A customer should read the public record as a starting point for due diligence: useful, specific and promising in places, but still thin where recovery evidence matters most.

DC01VOL gives a physical anchor, not a full capacity audit

Friktoria's own DC01VOL facility page is the clearest physical claim. It identifies the site as DC01VOL at Volos, Greece and gives a capacity figure of 350 square metres. It says construction follows international standards and lists several telecommunications and data-centre references, including ANSI/TIA/EIA-568 B2, ANSI/TIA/EIA-569 B, ANSI/TIA/EIS-606 B and ANSI/TIA/EIA-942 B. The same page says the site uses clean uninterrupted power from the public electric grid, renewable energy, three-phase diesel generating sets, UPS clusters and two different power supply lines to racks. It also lists temperature and humidity ranges, dust and foreign-entity protection, CCTV, access control, intrusion alarm and fire safety systems.

Those details matter because they locate the company's marketed capacity in a specific type of asset. This is not a hyperscale campus and not a national cloud region. The evidence points to a compact data-centre room or facility serving colocation, dedicated server and VPS customers from Volos. A 350 square metre site can be commercially meaningful for local businesses, web agencies, software firms, regional organisations and customers that want Greek hosting without buying from a larger Athens or international platform. It can also become a single point of failure if the same physical room carries too many customer assumptions.

The 350 square metre figure should not be confused with usable IT load. Floor area is an envelope. Usable capacity depends on rack density, power delivery, cooling headroom, cable paths, reserved space, fire zones, maintenance access and how much of the room is equipped. Friktoria's public page does not disclose total contracted utility capacity, UPS capacity, generator capacity, generator fuel storage, live load, spare power headroom, cooling capacity in kilowatts, number of cabinets in service, or occupancy.

It also does not disclose whether the 350 square metres are all technical white space or a wider facility area that includes office, support and plant zones.

The company's images on that page show racks, network equipment, power equipment, cooling equipment and office areas, which strengthens the reality signal. Images are still not an audit. They can show that equipment exists and that the operator has photographed a working environment; they cannot prove that all racks are powered by independent A/B paths, that cooling survives a unit failure, that fire systems are tested under load, or that the room is maintained under the same state today as the photos imply. The date and scope of the images are not independently fixed in the public record.

The standards references need the same caution. Friktoria's page lists telecommunications and data-centre standards, but an operator's statement that construction follows or uses standards is not the same as a current TIA-942 certification listing or a third-party test report. Friktoria's public page says "Tier-III Design Data Center" on the colocation page, but the reviewed public record does not show a certificate number, certification body, scope statement or audit date.

The fair reading is therefore: DC01VOL is a credible physical anchor for a small Greek hosting and colocation provider, but public evidence does not yet translate the room into audited, fully quantified data-centre capacity. Customers should ask for the current as-built package, not only the public description.

Retail colocation reveals the scale of the operating surface

The Friktoria colocation catalogue helps size the commercial offer. It sells 1U and 5U services in a shared 19-inch rack, 10U and 21U portions, full-rack hosting and a floor-footprint option for a customer-provided rack. The lower packages include one 10/100/1000 Mbps Ethernet port, one IPv4 address and single-phase 230VAC power through a common A or B 16A connection. The full rack includes a 400/230VAC three-phase socket via dedicated 3x16A power. The footprint option similarly provides a 60 cm by 100 cm space, a gigabit-capable switch port, one IPv4 address and a dedicated 3x16A three-phase connection described as 11 kVA.

This is useful because it makes the marketed capacity tangible. Friktoria is not only selling abstract cloud language. It is selling rack units, rack portions, a full rack, power sockets, switch ports, IPv4 addresses and remote-hands minutes. That is the language of a small colocation provider. It also implies operational obligations: access control, equipment receipt, rack and stack, power provisioning, switch-port management, IP assignment, remote-hands response, ticket handling, maintenance notice, and recovery after a fault.

The same catalogue also shows capacity constraints. A one-IPv4-address default per package is consistent with a small address pool. The full visible AS204932 route is one /24, so every public IPv4 assignment matters. A gigabit-capable Ethernet port in a package does not mean the external network has enough upstream capacity to deliver full-rate traffic for many customers simultaneously. The facility page says external links exceed 400 MBps; even assuming the intended unit means Mbps or another practical bandwidth measure, the public route data still points to a compact edge, not a large carrier-neutral interconnection node.

The dedicated-server and VPS pages reinforce that picture. The dedicated-server page advertises Greek, European and US regions, with several Greek-region server configurations, older Intel desktop and Xeon hardware, variable RAM and disk choices, and low monthly prices. The VPS page advertises high-availability KVM VPS products with 32 Mbps ports and unlimited traffic. Those are plausible small-provider products. They also create a question: which services are actually hosted in DC01VOL, which are sourced from other Greek, European or US facilities, and which resilience claims apply only to the Volos infrastructure rather than to resale or remote capacity.

This operator boundary should be explicit. A customer buying colocation from DC01VOL is exposed to the Volos room, its utility connection, cooling plant, fire systems, security, local staff and carrier handoffs. A customer buying a Greek dedicated server may be exposed to the same room, but the product page alone does not prove it. A customer buying a European or US dedicated server may be using supplier infrastructure outside Friktoria's direct physical control. A customer buying a VPS may depend on Friktoria's virtualization cluster, storage layer, internal switching and upstream capacity rather than a single physical server.

The public pages do not provide a service-by-service location and failure-domain map.

The retail catalogue therefore strengthens the evidence that Friktoria sells real hosted-infrastructure services. It also shows why the capacity claim has to be parsed by product. "Data-centre capacity" can mean a floor footprint with 11 kVA, a full rack with three-phase power, a 1U slot on common A or B power, a dedicated server, a VPS, or an outsourced location. Each version has a different failure path.

The power story is detailed, but the hard proof is still missing

Power is the central dependency for this profile. Friktoria's electrical power distribution page gives more detail than many small hosting sites. It says the facility takes primary electricity from the wider public grid, naming DEDDIE for Greece; says consumed grid power is backed by renewable-source agreements; says the site supplements grid supply with stand-alone diesel generators; describes main distribution boards feeding UPS systems, load banks and cooling; and says Friktoria operates two different UPS clusters for two power supply lines, A and B, to routers, switches, racks, servers and other equipment.

The same page states that the UPS systems can provide power to servers and breakers for up to ten minutes, giving enough time to start a backup diesel generator after an outage. That is useful specificity. It gives customers a bridge interval to test against. It also exposes the remaining questions. What is the generator start success rate? How often is it tested under live or load-bank conditions? How much diesel is stored on site? How quickly can fuel be replenished during a regional disruption? Are both A and B UPS clusters fed by independent paths all the way to each rack, or are there shared upstream components?

Are the lower 1U and 5U packages on common A or B power rather than dual-cord A plus B? What happens to single-cord customer devices?

The wider grid context matters because Friktoria's public page names DEDDIE. The Hellenic Electricity Distribution Network Operator page describes HEDNO/DEDDIE as responsible for operation, maintenance and development of Greece's electricity distribution network. The IPTO about page describes IPTO as owner and operator of the Hellenic Electricity Transmission System, with a mission around safe, efficient and reliable electricity supply. For a Volos facility, the customer question is not simply whether Greece has a grid operator. It is where the facility connects, whether it has one or more utility feeds, how local distribution faults are handled, and how the building rides through both short and extended outages.

The public record does not answer those questions. Friktoria says grid, renewable power, diesel generation, UPS clusters and A/B lines. It does not publish a single-line diagram, utility-feed independence, generator count, generator rating, fuel runtime, UPS battery condition, maintenance schedule, transfer-switch test history, or the result of a recent outage. A ten-minute UPS bridge can be appropriate when generator start is reliable; it can be inadequate if the generator fails to start, fuel delivery is delayed, or a transfer component is the shared weak point.

The colocation catalogue adds another nuance. Some small packages receive one 230VAC socket via common 16A power connections marked A or B, while larger packages move toward dedicated single-phase or three-phase power. If a customer's server has a single power supply and is connected to only one side, the existence of A and B infrastructure does not by itself protect that device. If the customer has dual power supplies but buys a package that only includes one socket, resilience depends on an upgraded design. A/B power is not a slogan; it is a service configuration.

The conclusion is not that Friktoria's power story is false. The conclusion is that public power evidence is self-attested and incomplete. The company has published enough to make a serious conversation possible. It has not published enough for a critical customer to accept the 99.999% nominal uptime statement without private verification.

Cooling, fire and room operations are the quiet constraints

Cooling often decides whether a small data room can sustain its marketed density. Friktoria's facility page states temperature ranges of 22 to 26 degrees Celsius, humidity from 10% to 80%, and dust and foreign-entity protection. It also includes cooling images. Those are reasonable public signals for a hosted-infrastructure room. They do not quantify cooling redundancy, capacity per rack, containment design, compressor or CRAC redundancy, maintenance bypass, alarm thresholds, water or refrigerant dependencies, or what happens if a cooling unit fails during a heat wave.

That gap matters because Friktoria's product range spans low-cost VPS, dedicated servers and colocation packages up to a full rack and a floor footprint. A customer installing one modern high-density rack can consume more power and cooling headroom than several older low-power servers. The dedicated-server catalogue includes a mix of older desktop CPUs, older Xeons and newer NVMe options. The physical diversity of customer and provider equipment means the room may not have one uniform risk profile. Some cabinets may be modest; others may become hot spots.

Fire and security claims need similar discipline. The facility page lists CCTV, dual video recording, people access control, intrusion alarm, fire safety and fire alarm systems. Those are basic ingredients of a responsible facility. The public record does not identify the fire detection technology, suppression method, zoning, response procedures, post-event restoration plan, insurance requirements or customer notification terms. It also does not state whether the systems have independent inspection records or whether any certification covers the current installation.

Operational staffing is also unresolved. The colocation packages include five minutes of remote-hands service, which suggests Friktoria expects customers to need occasional physical assistance. The page does not publish staffing hours, on-call response times, escalation paths, spares availability, change windows, customer-access rules or maintenance notice terms. For a small facility, those details can matter as much as the plant design. A perfectly adequate room can still become risky if a fault occurs when the right engineer is unavailable or when customer communication is slow.

The operating-status hypothesis for this slot was a thin public footprint, and the cooling, fire and staffing evidence confirms that caution. There is enough visible material to say the operator has a facility narrative. There is not enough to say the facility has independently demonstrated resilience under heat, fire-alarm, maintenance or staffing stress. The public standard should therefore be "specific but unproven," not "generic" and not "fully verified."

AS204932 proves reachability, not broad internet capacity

The network evidence is stronger because it can be measured from public routing systems. RIPEstat's AS overview lists the holder as FriktoriaNet George Sgouridis trading as Friktoria.com - Data Center Services. and marks AS204932 announced on 12 July 2026. RIPE announced-prefix data shows 185.246.164.0/24 as the current announced prefix. RIPE routing status reports one IPv4 prefix, 256 IPv4 addresses, no IPv6 /48 equivalents and two observed neighbours.

The address registration is consistent with a data-centre label. The RDAP IP record for 185.246.164.0 identifies 185.246.164.0 through 185.246.164.255 as DATA-CENTER, ASSIGNED PA, country Greece, with George Sgouridis in administrative and technical roles. RIPE address-space hierarchy shows that exact /24 as DATA-CENTER and the less-specific 185.246.164.0 to 185.246.167.255 allocation as GR-FRIKTORIACOM-20180215. That supports the reading that the routed /24 is not random transit space; it is the operator's own Greek address surface.

The scale remains small. One /24 is enough for management, provider services, colocation defaults and a modest hosted-server business. It is not enough to prove large cloud scale, broad customer population, or spare public IPv4 abundance. A colocation customer can use its own space or a carrier's assignment, so a /24 does not cap the facility. It does cap what can be inferred from Friktoria's own visible routes.

RPKI also needs caution. The RIPE RPKI validation query returns unknown for AS204932 and 185.246.164.0/24, with no validating ROAs in that view. Unknown is not the same as invalid, and it does not mean the route is unreachable. It does mean the public record does not show a positive RPKI origin validation signal for the one visible IPv4 route. For customers that care about route security hygiene, that is a fixable but relevant gap.

IPv6 is another gap. RIPE routing consistency lists an IPv6 prefix, 2a0d:8880::/96, in registered data but not in BGP in the reviewed snapshot. RIPE WHOIS for the broader IPv6 allocation points to GR-FRIKTORIACOM-20180215 and ORG-GSTA2-RIPE, but public routing status sees no active IPv6 announcement for the AS. That may reflect an unused allocation, a registration artifact, a low-visibility path or a deliberate IPv4-only production edge. Public evidence supports the narrow statement: Friktoria has an IPv6 registration context, but no visible active IPv6 route in the current RIPE routing status.

The independent route summaries are broadly aligned. CAIDA AS Rank marks AS204932 as seen, with one prefix, 256 addresses and two provider relationships. BGP.Tools and Hurricane Electric's BGP Toolkit are useful cross-checks for the same public edge, even where the detailed measurement comes from RIPE and CAIDA. A RIPE routing-history query also keeps the same /24 tied to AS204932 across the prior-year query window, with varying peer visibility. The picture is stable enough to call AS204932 real and current. It is too small to call broad or heavily redundant.

Carrier diversity exists in BGP, but it is still a narrow edge

The carrier picture is better than a single-upstream microhost, but not yet strong enough for a critical resilience claim. RIPE ASN-neighbours data shows two observed neighbours in the latest snapshot: AS1241 and AS3329. RIPE's AS overview for AS1241 identifies AS1241 as FORTHNET-GR Nova Telecommunications & Media Single Member S.A. RIPE's AS overview for AS3329 identifies AS3329 as HOL-GR Vodafone-panafon Hellenic Telecommunications Company SA. That means Friktoria's current public BGP edge is not visibly dependent on only one upstream in the RIPE view.

The registered policy is wider in one direction and narrower in another. RIPE WHOIS for AS204932 lists import and export entries for AS12361, AS3329 and AS1241. The RIPE routing-consistency view marks AS3329 and AS1241 as present in both WHOIS and BGP, but AS12361 as present in WHOIS and not in BGP. RIPE's AS overview for AS12361 identifies that AS as Vodafone-panafon Hellenic Telecommunications Company SA and marks it not announced in the queried view. The public reading should therefore be current two-provider visibility, not three-provider live diversity.

Two observed upstreams are meaningful. If properly engineered, they can provide route failover when one upstream has a fault. Yet public BGP cannot show physical entrance diversity, fibre path separation, meet-me-room design, switch redundancy, upstream contract terms, local-loop diversity or whether both carriers enter through the same building path. It also cannot show whether customer colocation circuits can bypass Friktoria's own routed edge or whether all customers default to the same switches and upstreams.

Carrier names on the facility page add context but not proof. Friktoria lists Vodafone, HOL, FORTHnet, HellasNet and a metropolitan microwave network among internet providers. The observed BGP neighbours correspond to the Nova/Forthnet and Vodafone/HOL side of that list. HellasNet and the microwave reference are not established as current active upstreams by the public AS204932 routing snapshot. They may be historical, physical, service-specific or otherwise outside public BGP. A customer should not count them as live failover providers without current evidence.

The network edge also raises capacity questions. A package with one gigabit-capable switch port can be useful for a server, but it does not prove gigabit upstream capacity for every customer. The visible AS carries one /24. The facility page says external links exceed 400 MBps. The BGP edge shows two providers. None of those facts publishes committed information rate, burst policy, transit capacity, packet-loss history, DDoS mitigation, route filtering, maintenance windows or customer failover results. Those details decide whether "two upstreams" becomes resilience or merely a configuration line.

The right grade for carrier evidence is medium at the network-identity layer and weak-to-medium at the facility-diversity layer. The ASN is real. The /24 is announced. Two upstreams are seen. But the public evidence does not yet prove path diversity inside the facility or customer-level survival during a carrier-meet interruption.

Failure paths start where marketing language becomes an operating event

The main failure path is a utility outage. Friktoria says it uses the public grid, UPS clusters and diesel generating sets. The customer's concern is the full chain: grid drop, UPS carry time, generator start, transfer, generator load, fuel endurance, return to grid and post-event inspection. A ten-minute UPS bridge gives a clear time window. It also makes generator-start reliability and fuel supply decisive. If the generator starts cleanly, ten minutes may be enough. If it does not, the room can exhaust battery time quickly.

The second path is an A or B power failure inside the room. Friktoria says it operates two UPS clusters for A and B lines. But package details show that some smaller offerings provide one socket via common A or B power. A single-cord server on one side is not protected by the presence of another side unless the customer has arranged dual input or an automatic transfer device. The public materials do not say how many customer devices are single-cord, whether dual-cord options are standard, or whether shared breaker loading is monitored against customer growth.

The third path is cooling. A utility outage tests generators and UPS systems; a cooling failure tests heat removal and operational response. The public material gives target temperature and humidity ranges but not redundancy design. If cooling capacity is only just sufficient during normal operation, a hot day or failed unit can force load shedding before power equipment becomes the limiting factor. Customers with high-density racks or storage-heavy dedicated servers should ask for rack-density limits, cooling redundancy and alarm response.

The fourth path is carrier interruption. The public edge currently has two observed upstreams. That is better than one, but customer survival depends on how those upstreams are implemented. If the same switch, router, optical path or building entry is common to both, an incident can still remove connectivity. If AS204932 is the default address source for hosted services, the one /24 and its route policy become a shared customer dependency. If customers bring their own carriers, the meet-me and cross-connect design become more important than Friktoria's own AS.

The fifth path is fire or physical access disruption. Fire alarm, suppression, security and access control appear in the facility description, but the public record does not describe event recovery. If a fire alarm triggers a shutdown, who authorises re-entry? How are customer devices inspected? How fast are customers told whether equipment is safe? How is remote-hands prioritised? These questions matter more in a small facility because customer workloads may rely on a few people and a few rooms.

The sixth path is capacity oversubscription. The catalogue offers low-cost dedicated servers and VPS plans alongside colocation. If retail demand grows faster than power, cooling, upstream and staff capacity, the marketed room can become constrained even without a dramatic incident. Public evidence does not show current occupancy, live power draw, transit utilisation, storage redundancy or support queue performance. That lack of evidence is not a finding of oversubscription. It is the reason not to assume spare capacity.

Who is affected if DC01VOL fails

The affected group is likely small and regional, but not trivial. A local company may use Friktoria for its web presence, mail, customer portal, internal application, backup server or accounting system. A web agency may host several client sites on a VPS or dedicated server. A small technology company may colocate a firewall, backup appliance, storage server or application node. A customer outside Greece may buy the service because it is inexpensive and geographically useful. In all cases, the customer's dependency can be larger than the provider's public footprint suggests.

The 1U and 5U offers are especially important because they attract customers who may not have a full infrastructure team. A buyer of one or five rack units may assume that the facility's A/B, generator and provider claims carry the full burden. In reality, the customer still has to design dual power supplies, backups, off-site replication, monitoring and recovery. A single server in a single Volos room is not made highly available by a nominal uptime number.

Dedicated-server and VPS customers face a different version of the same issue. They may never see the facility. They depend on Friktoria's hardware choices, virtualization layer, storage design, upstreams and support. The VPS page uses high-availability language, but public service pages do not show cluster topology, storage replication, failure-domain separation, backup policy or restoration targets. Customers should ask whether high availability means host-level migration, storage redundancy, network failover, or simply a marketed class of service.

Customers that treat Greek location as a resilience feature also need care. Local hosting can help with latency, jurisdiction, language, billing and support. It can also concentrate risk if the customer has no independent secondary site. If a business hosts production in DC01VOL and keeps backups in the same room or on the same provider edge, the local facility becomes a shared dependency. If it keeps backups with another provider, another city or another cloud, Friktoria can be a useful local node in a wider design.

The public article therefore should not overstate the hazard. Friktoria is a small provider, and most customers may buy services that match modest expectations. The point is proportionality. A small regional data-centre can be valuable when the customer knows what it is buying. It becomes risky when marketed uptime and capacity are treated as substitutes for evidence.

What evidence would move the grade

The first missing item is a current facility scope statement. Friktoria could materially improve public confidence by stating how much of the 350 square metres is technical space, how many cabinets are equipped, what rack-density range is supported, how much power is installed, how much is committed, and how much is reserved for redundancy. This does not require naming customers. It requires aligning capacity language with the actual operating envelope.

The second missing item is power test evidence. Customers should ask for generator count, generator rating, fuel capacity, fuel contracts, UPS battery autonomy under current load, transfer-switch testing, load-bank history, maintenance dates and a clear explanation of how A and B lines are delivered to different colocation products. The public power page is a useful start because it names UPS clusters and a ten-minute bridge interval. The next step is proving that those pieces work under load.

The third missing item is cooling and fire evidence. Customers should ask for cooling capacity, redundancy, cabinet-density limits, high-temperature alarm thresholds, suppression technology, inspection dates, fire-zone design and recovery steps after an alarm. A small facility can be resilient if it is disciplined. Public pages with photos and target temperature ranges do not prove that discipline under stress.

The fourth missing item is carrier documentation. The public BGP view supports two current observed upstreams, AS1241 and AS3329. Customers should ask whether those arrive by physically separate paths, whether they terminate on separate routers and switches, whether maintenance can be done without losing both, whether AS12361 is still a usable option, and whether customer cross-connects can use carriers other than Friktoria's default routed edge. The answer decides whether the network is merely multihomed or truly route-diverse for customers.

The fifth missing item is route-security hygiene. One visible IPv4 route with unknown RPKI validation is not a crisis, but it is a clean improvement area. A valid ROA for 185.246.164.0/24, current route objects, documented filtering and a public statement on IPv6 service status would strengthen the network evidence. If the registered IPv6 allocation is not intended for service, saying so would also reduce ambiguity.

The sixth missing item is customer recovery evidence. This can be anonymised. A short public statement saying that a generator test, upstream withdrawal, cooling failover or customer restoration exercise was completed on a given date would carry more weight than a nominal uptime percentage. Even a private customer diligence pack would be enough for serious buyers. The public record currently has no such material.

The evidence grade

Friktoria earns a medium network evidence grade and a weak-to-medium facility evidence grade. The network grade is medium because AS204932 is active, registered to the named Volos operator, announces 185.246.164.0/24, appears in current RIPE routing status, and has two observed upstreams. It is not strong because the public edge is one /24, no visible IPv6 route, unknown RPKI validation, no customer cone, no peer diversity and no published failover result.

The facility grade is lower because the strongest facility material comes from Friktoria itself. The operator publishes a specific DC01VOL description, power-distribution explanation, colocation catalogue, dedicated-server catalogue and VPS catalogue. Those pages are useful and more detailed than generic hosting copy. They still do not replace third-party certification, as-built disclosure, live-load figures, generator runtime evidence, cooling redundancy proof, carrier entrance diagrams or customer recovery records.

The practical conclusion is clear. FriktoriaNet George Sgouridis trading as Friktoria.com - Data Center Services. should be treated as a real small Greek data-centre and hosting operator with a live AS, a Volos facility claim and visible colocation products. It should not be treated as a fully proven resilient data-centre platform on the public record alone.

Customers can use the public information to frame the right questions: how much capacity is actually equipped, how power moves from grid to UPS to generator to rack, how cooling survives a unit failure, how two observed upstreams reach the building, and how a real customer is restored after a fault.

Until those questions are answered with current evidence, marketed capacity should remain a hypothesis. The operator's public footprint is enough to start diligence. It is not enough to end it.