Works now! Thanks
... Not sure why my latency is so high though. I'm ~13-16ms from downtown LA. I'm using fq_codel, cos the unifi firewall doesn't do cake
Nils,I've now fixed the issue where the initial latency spike skews baseline results by switching to using 75th percentile for baseline calculation. Thanks for catching that.Geoff,I may have fixed that now. Our Vultr VPS was hitting its bandwidth limit apparently, and they throttled it. Switched to hosting it locally in El Paso as fast as I could. Now tests can achieve saturation throughput.Dave,If you try it again does it still fail to ping currently?Thanks,RobertOn Sun, Jun 15, 2025 at 11:32 AM dave seddon <dave.seddon.ca@gmail.com> wrote:Very cool Frank!When I run it, it says all of my latency is 0.0ms. This is firefox-139.0.1 on NixOS unstable.I guess soem of the CORS headers are screwed up?Cross-Origin Request Blocked: The Same Origin Policy disallows reading the remote resource at https://test-elp.libreqos.com:8005/ping?cb=242578. (Reason: CORS request did not succeed). Status code: (null).I don't know, I guess you probably _do_ want these objects to be cacheable via the CDN?GET /ping?cb=819806 HTTP/1.1
Host: test-elp.libreqos.com:8005
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:139.0) Gecko/20100101 Firefox/139.0
Accept: */*
Accept-Language: en-US,en;q=0.5
Accept-Encoding: gzip, deflate, br, zstd
Pragma: no-cache
Cache-Control: no-store, no-cache, must-revalidate
X-Priority: high
X-Ping-Attempt: 0
Referer: https://test.libreqos.com/
Origin: https://test.libreqos.com
Connection: keep-alive
Sec-Fetch-Dest: empty
Sec-Fetch-Mode: cors
Sec-Fetch-Site: same-siteHTTP/1.1 200 OK
date: Sun, 15 Jun 2025 17:13:55 GMT
server: uvicorn
cache-control: no-store, no-cache, must-revalidate
pragma: no-cache
x-ping-server: dedicated
x-priority-processed: true
x-ping-timeouts-seen: 0
content-length: 4
content-type: text/plain; charset=utf-8
access-control-allow-origin: *
access-control-allow-credentials: trueYour Nginx server also can have caching enabled for the favorite icon, and the javascript, which should make it faster to load.HTTP/2 200
server: nginx/1.18.0 (Ubuntu)
date: Sun, 15 Jun 2025 17:13:35 GMT
content-type: text/javascript; charset=utf-8
vary: Accept-Encoding
last-modified: Tue, 10 Jun 2025 04:28:53 GMT
strict-transport-security: max-age=63072000
x-content-type-options: nosniff
x-frame-options: SAMEORIGIN
content-encoding: gzip
X-Firefox-Spdy: h2I've found that you can set the caching for the CORS "preflight" requests too and it definitely helps. e.g. access-control-max-age 345600On Sun, Jun 15, 2025 at 9:53 AM Nils Andreas Svee via Cake <cake@lists.bufferbloat.net> wrote:_______________________________________________Just tried it too: https://imgur.com/a/F2cNZd4, also getting A+ overall, and "only" A for bidirectional.I seem to be getting a 200+ ms spike right at the beginning of the baseline test, which skews the results.This happened at least on a couple of tests, but after a few tries it got better. At least with ICMP ping separately I don't see any spikes like that.Best RegardsNilsOn Sun, Jun 15, 2025, at 23:20, the keyboard of geoff goodfellow via Bloat wrote:re: thoughts and feedback!your https://test.libreqos.com tests of [Single User Test] & [Virtual Household Mode] both give yours truly bufferbloat grades of A+whereas the https://www.waveform.com/tools/bufferbloat test gives yours truly a bufferbloat brade grade of C#1.) Why/What's the difference?#2.) Who/Which one to believe¿gOn Sun, Jun 15, 2025 at 4:58 AM Frantisek Borsik via Starlink <starlink@lists.bufferbloat.net> wrote:Hello to all,We're excited to announce the release of the LibreQoS Bufferbloat Test – an open-source bufferbloat testing solution designed specifically for ISPs and network operators to deploy for their customers.LinkWhat Makes This DifferentWhile there are several bufferbloat testing tools available, this platform addresses a critical gap: ISP-deployable infrastructure that provides both traditional testing and realistic household simulation.As Dave Täht highlighted in his influential article "What's Wrong with Speed Tests", traditional speed tests fail to measure what users actually experience. We tried to address Dave's points to make a better speed test that focuses on the metric that matters: latency under load in realistic usage scenarios.Two Complementary Test ModesSingle User Test Mode
- Traditional sequential load testing (baseline → download → upload → bidirectional)
- Measures working latency and jitter during each phase
- Familiar A+ to F grading based on latency under load increases
- Comparable to existing tools like DSLReports Speed Test and Waveform Bufferbloat Test
Virtual Household Mode (The Innovation)Process-isolated simulation of 4 concurrent users with authentic traffic patterns:
- Alex (Gaming): 1.5 Mbps constant, jitter-sensitive for competitive gaming
- Sarah (Video Conference): 2.5 Mbps bidirectional, Teams simulation with working latency monitoring
- Jake (Netflix HD): 25 Mbps bursts (1s on, 4s off), realistic streaming patterns
- Computer (Background): Up to 200 Mbps continuous download, system updates
Real-world relevance: Tests latency under load when multiple family members are online simultaneouslyAdvanced grading: Network fairness, jitter measurement, and per-user working latency analysisWhy (not only) ISPs Need ThisThe traditional approach of sending customers to third-party speed test sites has limitations:
- No control over test methodology or server placement
- Limited correlation with customer support tickets
- Generic results that don't reflect real-world usage patterns
- No integration with ISP operational systems
This platform enables (not only) ISPs to:
- Host their own testing infrastructure with full control
- Integrate with support systems via telemetry APIs
- Provide customers with realistic household testing scenarios
- Correlate test results with network performance and customer complaints
Open Source & CommunityThe entire platform is open source and available here: https://github.com/LibreQoE/bufferbloat_testWe've designed this to be:
- Easy to deploy for (not only) ISPs of any size
- Scientifically meaningful in its measurement methodology
- Realistic in its simulation of actual household usage
- Integrable with existing ISP operational workflows
Community Feedback RequestedWe'd love feedback from the bufferbloat.net community on:
- Test methodology: Are we measuring the right metrics?
- Grading thresholds: Do our A+ to F grades align with real-world impact?
- Virtual household scenarios: What other realistic usage patterns should we simulate?
- ISP adoption: What barriers exist for ISP deployment?
Technical DiscussionWe'd welcome discussion about:
- Measurement accuracy for working latency and jitter in virtual household mode
- Traffic pattern authenticity (gaming, video conferencing, streaming)
- Grading methodology for latency under load in complex multi-user scenarios
- Integration approaches for ISP operational systems
The platform represents our attempt to bridge the gap between academic bufferbloat research and practical ISP operations, building on the foundational work of researchers like Dave Täht and the broader bufferbloat community. We believe that widespread ISP deployment of proper bufferbloat testing infrastructure will ultimately benefit the entire internet ecosystem.Looking forward to the community's thoughts and feedback!Best regards,The LibreQoS Team
In loving memory of Dave Täht: 1965-2025
https://libreqos.io/2025/04/01/in-loving-memory-of-dave/
https://www.linkedin.com/in/frantisekborsik
Signal, Telegram, WhatsApp: +421919416714
iMessage, mobile: +420775230885
Skype: casioa5302ca
_______________________________________________Starlink mailing list--living as The Truth is True_______________________________________________Bloat mailing list
Cake mailing list
Cake@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/cake
--Regards,Dave Seddon
+1 415 857 5102
--
