Internet weather · Is it down? · Access Health
Is it the internet, is it a service,or is it your access network?
- 01Internet weathershared path or your own path
- 02Is it down?64 popular services, own baseline
- 03Access Healthoutage radar, vendor scorecard
- 04Public status pageanonymous, launching now
- 05Plan-scaled protectionin Session Info
shared path, not you
its own normal
customer-minutes lost
“all fine”
00The problem: one phone call, three suspects
"The internet is down" can mean three very different things, and each one needs a different team and a different answer to the customer.
A shared path everyone uses
A slow cache, a congested exchange or a transit problem hits many networks at once. Nothing in your building can fix it, yet your support desk takes the calls.
A popular app having trouble
A payment app or a game fails nationally. Customers blame their ISP first, and support learns about it from the queue, not from a dashboard.
Your own fiber, OLTs and VLANs
A cut fiber or a lost OLT power feed drops a whole area. Session logs show hundreds of disconnects, but nobody sees them as one event with a cause.
01Internet weather: is it the internet, or is it us?
NOC2 combines what every gateway sees, per destination network owner (Google, Meta, Akamai, Amazon, Microsoft, Jio and more) and per city, into a weather map. Shared weather means a shared path. Weather over only one operator means that operator's own path. Click a square.
Likely the CDN, IX or transit, not your network.
"Your Google answer time is slower than most operators measured in your region."
Shown only when at least 3 other operators measure the same thing; otherwise it rolls up to the region. It never names anyone, and each operator sees only its own gateways by name.
What a subscriber in that city can do right now
ESTIMATEThroughput is estimated from round-trip time and loss (the Mathis formula), and labelled as an estimate everywhere it appears.
Where your users could be faster, once a month
- Caches bypassed: traffic that should be served close by but is not.
- Peering missing: owners you reach the long way round.
- Slow DNS upstreams: the resolvers that add time to every page.
02“Is it down?” for popular services
The gateway resolves each service's domains itself, then counts new connections and how fast that service's servers answer, measured at the gateway, so the subscriber's own line is not included. Each service is judged only against its own normal on that gateway: the same hour on past days, else the last 7 days. Never against a fixed number.
Demo values are illustrative. The shaded band is the service's own normal for that hour; the line is today.
Can be perfectly healthy
Some services have only 66% of connections answered on a normal day, because remote caches answer slowly. A fixed threshold would cry wolf all day; NOC2 compares each service with itself.
Operators at the same time
Down at 3 or more operators at once: "the service is having trouble". Down at only one operator: "that operator's path". Support knows which answer to give before the queue fills.
Grey until it knows
A new gateway learns what is normal. Meanwhile live readings show in grey: "WhatsApp: learning normal · measured now 96.6% answered · ~1 ms". Nothing turns green or red before it has its own baseline.
A starter catalog of 64 popular services
03Access Health: your own access network, as events
A new NOC2 menu for the part of the network only you own: fiber, OLTs, VLANs and the routers in your customers' homes.
10 sessions, 60 seconds, half the VLAN
10 or more sessions on one VLAN drop within 60 s for Lost-Carrier or Idle-Timeout, and they are at least 50% of that VLAN. Operator resets are shown apart and never counted.
"OLT Block-C lost 86 of 86 sessions"
Name each VLAN after its OLT or area once. Outages and incidents then speak your field team's language, not "VLAN 2417".
The lines that never quite heal
Lines that were faulty for at least 50% of the last 7 days, per gateway: the list for a planned visit, before the customer churns.
Router vendor scorecard: is it the line or the Wi-Fi?
Per CPE vendor (labelled "vendor, not model", from the MAC OUI). Line delay (floor) is the line itself; extra delay on top of it is buffering or Wi-Fi in the home. Click a bar.
Illustrative values. Groups only from 20 subscribers up; randomised MACs form their own group; fewer than 20 measured subscribers is a hint, never a verdict.
04A public internet status page your customers can trust
An anonymous status page for end users and the press, launching now at status.bngsoft.com. It is still learning, so today it honestly says "Not enough data" until enough operators have a baseline.
Mock-up of the launch state. Layout may change as the page learns.
≥3 operators, and at least half
A problem shows only when 3 or more different operators see it and they are at least half of the operators with data there.
Never a name, never a count
No operator names, no counts of anything, no tracking. Cities only from a fixed metro list; a wide outage reads "across India" or "parts of India", so no one's footprint can be mapped.
"Not enough data", never "all fine"
Each operator can opt out. An invisible, self-hosted bot check keeps it clean, with no Google or Cloudflare captcha.
05Session Info: protection that scales with the plan
Upload protection thresholds now scale with the subscriber's upload plan: ×1 per 100 Mbps, from ×1 to ×10. Session Info shows the scale, e.g. "plan scale ×3". Fixed limits sized for 100M homes were pushing 300M–1G business customers into LIMITED or BLOCKED, and their calls dropped. After the change: zero false escalations.
06Built to be believed
Learning, not enough data, not measured: each says so, in grey. A green square always means measured and normal.
Every cross-operator fact is anonymous. Each operator sees only its own gateways by name.
VLAN flags, upstream findings, slow destinations, DDoS and outages merge into one incident with a likely cause and the evidence.
07What operators get
Stop guessing
Know whether it's the internet or you, per owner and per city, before anyone escalates.
Proof, not opinion
"It's not your network" backed by a rule your customers and your management can read.
Ahead of the queue
Know a payment app is failing nationally before customers flood support.
One event per cut
See an OLT power cut as one event with a likely cause and customer-minutes lost.
Line or Wi-Fi?
Learn which CPE vendor makes the complaints, and whether it is the line or the home.
A page to trust
Give your customers a public status page that never pretends to know.
Ask it the next time someone says "the internet is down".
We will walk through Internet Observatory 3 on a live NOC2: the weather map, "Is it down?", the outage radar and the vendor scorecard, on your questions.
Examples on this page are illustrative and anonymised. Gateway names, VLAN numbers, cities, times and values are invented; operators are shown only as anonymous letters. Service and company names (including Google, Meta, Akamai, Amazon, Microsoft, Jio, UPI and the apps listed) are trademarks of their respective owners, named only as destinations subscribers reach; BNGSOFT is not affiliated with them. Throughput figures marked ESTIMATE are calculated, not measured.
Product page: Support & operations · Talk to us