Try off-hours
Russian QoS rules often soften at night (00:00-08:00 MSK). If your test now shows heavy asymmetry, run it again at 03:00 — if symmetric returns, the shaping is time-of-day-based. Confirms it's shaping, not infrastructure capacity.
Streams 10 MB over two paths and compares throughput. A gap is worth a look but proves little on its own: Cloudflare is hundreds of edges near you, our endpoint is a single server, and it is almost always the slower of the two.
Test downloads ~20 MB total (10 MB per path). Skip if you are on metered mobile data.
Cloudflare anycast lands on the edge closest to you; our endpoint is one server, the one serving this site. Our path is longer by definition and the gap never runs in our favour. What is worth reading is the size of the gap, not its existence — and if you are connected through our tunnel, this measures our own route rather than your ISP's network.
Asymmetric throughput suggests traffic shaping on the slower path. Symmetric-but-slow indicates baseline bandwidth limits. Two non-network factors are worth ruling out first.
Russian QoS rules often soften at night (00:00-08:00 MSK). If your test now shows heavy asymmetry, run it again at 03:00 — if symmetric returns, the shaping is time-of-day-based. Confirms it's shaping, not infrastructure capacity.
Mobile carriers (Tele2, Yota, T2) often pass traffic Rostelecom rate-limits. Tethering from a mobile hotspot is a quick A/B test. Long term: a mobile line as fallback covers moments when wired path degrades.
Throttling — not blocking — is the QoS policy where some flows get less bandwidth than others. The decision happens per-flow at routing boxes (the same TSPU stack that handles blocks). The discriminator varies: destination ASN (any Google CDN IP is rate-limited), SNI in the TLS ClientHello (any handshake to googlevideo.com), or behavioral patterns (long-lived large-payload connection consistent with video streaming). The user sees: YouTube is slow but Cloudflare is fast, both on the same network. This tool's two-path test surfaces shaping by comparing throughput through different paths. Symmetric = no shaping. Asymmetric = shaping on the slower path. The verdict doesn't tell you who's shaping (could be ISP, regulator, transit provider) — only that someone is.
Each test answers a different question. Together they make a full picture of your connection.
Check connectivity to popular services (YouTube, Discord, Telegram, and more) across multiple protocol layers.
Does QUIC (HTTP/3) actually reach the major CDNs from your network? Detects whether your browser falls back to HTTP/2.
What our server sees about you: IP, geolocation, every HTTP header your client sent. The exact picture any site receives.
Shaping is applied to what got recognised — a plan limit would slow both paths equally. Logrus hides the protocol, so there is nothing left to recognise, and works through transports until one gets through.
If it is not for you, do nothing: the access simply expires.
Technical tool for diagnosing the quality of a network connection. Does not provide means of access to information resources whose access is restricted under applicable law.