Jekyll2024-03-18T22:19:10+00:00/feed.xmlM-LabContributing to M-Lab’s Infrastruture is Easier Than Ever2024-03-18T00:00:00+00:002024-03-18T00:00:00+00:00/blog/contributing-mlab-infrastructure<p>We’ve added a new, flexible set of options to expand and diversify our global fleet of measurement vantage points.
<img src="/images/blog/2024-03-infrastructure/infra-options-diagram.png" alt="Diagram of Infrastructure Options" width="750" style="display:block; margin-left:auto; margin-right:auto" />
<!--more--></p>
<h2 id="motivations">Motivations</h2>
<p>Since M-Lab was founded in 2008, our mission has been to measure the Internet, save the data and make it universally accessible and useful. As the Internet evolves, we are committed to producing data and tools that reflect these evolutions. Our platform, a global fleet of virtual and physical servers, is a fundamental part of how we accomplish this.</p>
<p>In 2022, we announced our initiative <a href="https://www.measurementlab.net/blog/2022-mlab-to-the-cloud/">M-Lab to the Cloud</a>, our effort to expand our presence in cloud networks. Since that announcement, 35% of our traffic is directed to virtual servers.</p>
<p><img src="/images/blog/2024-03-infrastructure/virtual-traffic.png" alt="Virtual Server Traffic" width="750" style="display:block; margin-left:auto; margin-right:auto" /></p>
<p>As a continuation of those efforts, we’ve added a new, flexible set of options for contributing to M-Lab’s infrastructure.<strong>Our goal is to expand and diversify our platform to include more user-relevant pathways.</strong></p>
<h2 id="a-new-flexible-set-of-options">A new, flexible set of options</h2>
<p>New vantage points can be managed by M-Lab or managed by their host. They can be virtual or physical, and scaled to support either a full site or a minimal site. All options can be placed in “on-net” or “off-net” networks, or in other words, located in edge, access or interconnection points.</p>
<table>
<tr>
<td><strong>Deployment Option</strong>
</td>
<td><strong>Summary</strong>
</td>
<td><strong>Key Feature</strong>
</td>
<td><strong>Timeline</strong>
</td>
</tr>
<tr>
<td>Virtual, cloud deployments
</td>
<td>Sponsored or directly allocated cloud resources, managed by M-Lab
</td>
<td>Easiest option for cloud providers or organizations without direct access to a network.
</td>
<td>Available today
</td>
</tr>
<tr>
<td>Full Site
</td>
<td>Multiple colocated servers, managed by M-Lab
</td>
<td>Redundant support fit for regions with a higher testing demand
</td>
<td>Available today
</td>
</tr>
<tr>
<td>Minimal Site
</td>
<td>Single server, managed by M-Lab
</td>
<td>A lighter set of requirements for contributing physical, M-Lab managed servers
</td>
<td>We are open to Trusted Testers in 2024 Q2 and plan to make this option generally available by 2024 Q3.
</td>
</tr>
<tr>
<td>Host Managed
</td>
<td>Single server, running M-Lab software, managed by the host organization
</td>
<td>Make it easy for those already managing servers to contribute to the platform
</td>
<td>We are open to Trusted Testers in 2024 Q3 and plan to make this option generally available by 2024 Q4.
</td>
</tr>
</table>
<p>For a full overview of specifications, please see our <a href="https://www.measurementlab.net/contribute/#host-or-sponsor-an-m-lab-measurement-site">Contribute</a> page.</p>
<p>You can also <a href="https://youtu.be/Tm9Hmsv7jMA?feature=shared">watch</a> and <a href="https://bit.ly/m-lab-community-call-2024-03">review the slides</a> from our March 2024 Community Call, where we reviewed each option in detail.</p>
<h2 id="interested">Interested?</h2>
<p>If you are interested in contributing to M-Lab’s infrastructure, please fill out our <a href="https://docs.google.com/forms/d/e/1FAIpQLSe1wXKfQ0VIt_hZFatCwCaoOeeDpRv3JZDM_eAmIaksMuwB4g/viewform">Infrastructure Contribution Form</a> and a member of our team will reach out to you shortly.</p>Lai Yi Ohlsen, Nathan Kinkade, Stephen SolteszWe’ve added a new, flexible set of options to expand and diversify our global fleet of measurement vantage points.How should Internet quality be measured? Join us for our Community Call on February 1, 20242024-01-23T00:00:00+00:002024-01-23T00:00:00+00:00/blog/feb24-community-call<p>Access to high speed Internet does not necessarily mean access to high quality Internet. But if not speed then what? How <em>should</em> we measure Internet quality? Join our monthly community call on February 1, 2024 at 11am ET to discuss and contribute to our latest research initiative, sponsored by Internet Society Foundation. <!--more--></p>
<h2 id="introducing-internet-quality-barometer">Introducing Internet Quality Barometer</h2>
<p>Over the past decade, network and Internet measurement experts have been reaching a consensus that throughput measurements, more commonly referred to as “speed” tests, are not the only suitable proxy for the quality of an user’s experience using the Internet and that other metrics such as unloaded and loaded latency, jitter, and packet loss should also be considered. Additionally, experts often discuss the need for measurements that account for the impact of local wi-fi networks, as well as metadata to differentiate between different types of access technologies such as fiber and DSL. These characteristics are complex to consider when assessing the quality of an individual connection and only become more layered when analyzing the connectivity of a region in aggregate - for example, where should the data be collected from? How often? From which access technologies and using which measurement methodologies? And so on.</p>
<p>With support from the Internet Society Foundation, M-Lab is excited to be leading a research initiative that will consider all the relevant complexities in the context of their practical application within policy. Together with our diverse ecosystem of experts, we aim to create an Internet Quality Barometer (working title), a standardized approach to measuring Internet quality that decision-makers can use to make data-driven improvements to the Internet in their region.</p>
<h2 id="join-our-commmunity-call-to-contribute">Join our Commmunity Call to Contribute</h2>
<p>We invite you to join us on February 1, 2024 at 11am ET for our first monthly community call of 2024, where we will gather community feedback about who and what will make the research a success. Specifically:</p>
<ul>
<li>What datasets, tools and methodologies should be included?</li>
<li>What organizations, consortiums and experts should be consulted?</li>
<li>What conversations (e.g., newsletters, mailing lists etc.) should we be part of?</li>
<li>If you are interested in helping shape this work, how are you interested in contributing? Through an advisory committee? Through a survey? A workshop? Other formats?</li>
</ul>
<p>Can’t join the call but want to contribute or give feedback? Feel free to reach out to <a href="mailto:hello@measurementlab.net">hello@measurementlab.net</a>.</p>
<h2 id="registration">Registration</h2>
<p>Please <a href="https://us02web.zoom.us/meeting/register/tZwucuCgqTIiHdd1OVBDlvHVHKGfyGGoMn6U">register</a> for our community calls and reach out to <a href="mailto:hello@measurementlab.net">hello@measurementlab.net</a> if you have any questions.</p>Lai Yi OhlsenAccess to high speed Internet does not necessarily mean access to high quality Internet. But if not speed then what? How should we measure Internet quality? Join our monthly community call on February 1, 2024 at 11am ET to discuss and contribute to our latest research initiative, sponsored by Internet Society Foundation.2023 Year in Review2023-12-21T00:00:00+00:002023-12-21T00:00:00+00:00/blog/end-of-year-letter-2023<p>2023 was a big year! Here’s a summary of what we were up to and what’s coming up next. <!--more--></p>
<h2 id="executive-summary">Executive Summary</h2>
<p><a href="#year-in-review">In 2023 we</a>:</p>
<ul>
<li>Published three new datasets to BigQuery: Wehe, Reverse Traceroute and Cloudflare’s speed test and AIM data.</li>
<li>Expanded the virtual presence of the platform and developed MSAK, a tool for prototyping throughput performance measurements.</li>
<li>Began new partnerships with Cloudflare and Dioptra, hosted monthly Community Calls and presented M-Lab and the importance of open Internet Measurement data at 10+ conferences, gatherings and webinars.</li>
</ul>
<p><a href="#looking-ahead">In 2024 we plan to</a>:</p>
<ul>
<li>Increase the number of virtual sites and diversify cloud providers, pursue client integration partnerships and make it easier to contribute infrastructure to the M-Lab platform.</li>
<li>Enable calibrated measurements by increasing the number of client vantage points, redefine Internet Quality, and build out our Research and Technical Advisory Committee.</li>
</ul>
<p><a href="#get-involved">You can support M-Lab in the following ways</a>:</p>
<ul>
<li>Partner with us</li>
<li>Donate data or resources</li>
<li>Sponsor a site</li>
<li>Include our data and tools in your research</li>
</ul>
<h2 id="year-in-review">Year in Review</h2>
<p>Measurement & Data</p>
<p>We introduced the <a href="https://github.com/m-lab/autoloader">autoloading pipeline</a> in April with the announcement of <a href="https://www.measurementlab.net/blog/cloudflare-aimscoredata-announcement/">our partnership with Cloudflare</a>. Since then, the autoloading pipeline now publishes 10 new data types from Reverse Traceroute, Wehe, and Measurement Swiss Army Knife (MSAK) services, and the host environment.</p>
<ul>
<li><strong>Cloudflare’s speed test and AIM Data is now <a href="https://console.cloud.google.com/bigquery?project=measurement-lab&ws=!1m4!1m3!3m2!1smlab-cloudflare!2sspeedtest">published in BigQuery</a></strong>. Aggregated Internet Measurement (AIM) is a new format for displaying Internet quality in a way that makes sense to end users of the Internet while keeping the underlying data that network engineers need to make it better. Every Cloudflare speed test produces an AIM Score and each test result is published to BigQuery for longitudinal analysis at scale. You can read more about AIM <a href="https://blog.cloudflare.com/aim-database-for-internet-quality/">here</a>.</li>
<li><strong>Wehe data is now <a href="https://console.cloud.google.com/bigquery?project=measurement-lab&ws=!1m4!1m3!3m2!1smeasurement-lab!2swehe_raw">published in BigQuery</a></strong>.The Wehe application allows users to detect whether various applications are being throttled. Wehe has run on the M-Lab platform since 2020, and now the <a href="https://www.measurementlab.net/blog/wehe-bigquery-announcement/">data is published</a> via the autoloading pipeline. Learn more about Wehe from the October <a href="https://www.youtube.com/watch?v=fHDsBKle1lQ">community call</a>.</li>
<li><strong>More Reverse Traceroute data is being collected and <a href="https://console.cloud.google.com/bigquery?project=measurement-lab&ws=!1m4!1m3!3m2!1smeasurement-lab!2srevtr_raw">published to BigQuery.</a></strong>. Reverse traceroute, which constructs traceroutes from a client to an M-Lab server, has run on M-Lab for over a decade. Recently, M-Lab & the reverse traceroute team are generating reverse traceroutes for 1% of NDT measurements, which are then available in BigQuery. Combined with the traceroute datasets this provides forward and reverse path information for NDT tests. Look forward to announcements and more information soon.</li>
</ul>
<p>Platform</p>
<ul>
<li><strong>Over 30% of our traffic measures performance to Cloud servers.</strong> In the 15 years since M-Lab started, the Internet has changed significantly! In 2022, we started introducing testing to cloud servers to better measure performance aligned with the most common user experience. After completing a comparative analysis to ensure that there would be no adverse effects on the data, we began migrating an increasing amount of traffic. Learn more from our update at the September <a href="https://youtu.be/fcC2qSmMIy8?si=tA3ikfpVadoBjzip&t=1365">community call</a>.</li>
<li><strong>We started using Terraform for our automated Cloud management</strong>. Terraform, an infrastructure as code tool, will enable us to better support current and future partnerships.</li>
<li><strong>Measurement Swiss Army Knife (MSAK) <a href="https://www.measurementlab.net/blog/introducing-msak/">launched this year</a></strong>. This new measurement service provides multi-stream throughput and UDP latency measurements – the first open UDP latency measurement of its kind on the platform. The <a href="https://github.com/m-lab/msak/tree/main/cmd/msak-client">Go client</a> is available now, with the javascript client available soon. Learn more about MSAK from the July <a href="https://www.youtube.com/watch?v=Lvo-nlX5YeM&t=1s">community call</a>.</li>
</ul>
<p>Community Engagement</p>
<ul>
<li>We started a <strong>partnership with Dioptra,</strong> an Internet cartography research group within the Networks and Performance Analysis (NPA) team at the LIP6 computer science laboratory in Paris, France, to support our mutual efforts in open Internet measurement research.</li>
<li>Every month we hosted our <strong>community calls</strong> to share work from M-Lab, its partners, and the broader community — holding space to talk about changes in the Internet measurement ecosystem, provide updates from M-Lab, and get your input on M-Lab’s plans for the future. You can find recordings of every call on our <a href="https://www.youtube.com/channel/UCjNJYE8LXwUGhv1k8YWuOLQ">YouTube channel</a>.</li>
<li>We <strong>co-organized 2 community events</strong> including the Internet Measurement Conference Hackathon with <a href="https://ooni.org/">OONI</a>, <a href="https://pulse.internetsociety.org/">ISOC</a>, <a href="https://censoredplanet.org/">Censored Planet</a> and <a href="https://ioda.inetintel.cc.gatech.edu/">IODA</a> as well as the <a href="https://www.law.berkeley.edu/research/bclt/bcltevents/bridging-the-divide-answering-internet-policy-questions-with-cutting-edge-network-measurement-algorithms-datasets-and-platforms/">Bridging the Digital Divide</a> led by Arpit Gupta of UCSD and Tejas Narechania of UC Berkeley.</li>
<li>We were invited to present about M-Lab and open Internet Measurement data at <strong>10+ conferences, gatherings and webinars</strong> including <a href="https://www.newyorkfed.org/newsevents/events/regional_outreach/2023/0914-2023">New York Federal Reserve’s Northeast State Broadband Convening</a>, National Science Foundation’s <a href="https://wired23.github.io/">Long-term Research Directions in Wired Networking Community Workshop</a>, <a href="https://www.rightscon.org/program/#session-list">RightsCon</a>, Domos AI’s <a href="https://www.understandinglatency.com/">Understanding Latency</a>, ready.net’s’s <a href="https://summit.broadband.io/">Broadband Summit</a>, <a href="https://splintercon.net/">Splintercon</a>, <a href="https://www.eventbrite.com/e/open-data-dayweek-data-visualizing-global-internet-speed-quality-tickets-565562361247">Open Data Day</a>, <a href="https://africadigitalskillsconference.org/">Africa Digital Skills Conference</a> and more.</li>
</ul>
<p>Broadband Mapping</p>
<ul>
<li>Ready.net’s <a href="https://broadband.money/">broadband.money</a>, Breaking Point Solutions’ <a href="https://sites.google.com/site/breakingpointsolutionsllc/home">OptiExpress</a>, Exactly Labs <a href="https://broadbandmapping.com/explore?state=eyJnZW9zcGFjZU5hbWVzcGFjZSI6IkNPVU5USUVTIiwic3BlZWRUeXBlIjoiRG93bmxvYWQiLCJjYWxlbmRhclR5cGUiOiJUaGlzIHllYXIiLCJwcm92aWRlciI6eyJpZCI6IiIsIm9yZ2FuaXphdGlvbiI6IkFsbCBwcm92aWRlcnMiLCJhc24iOiIifSwic2VsZWN0ZWRHZW9zcGFjZSI6bnVsbCwic2VsZWN0ZWRTcGVlZEZpbHRlcnMiOlsiVU5TRVJWRUQiLCJVTkRFUlNFUlZFRCIsIlNFUlZFRCJdLCJ6b29tIjozLCJjZW50ZXIiOls0NS41NjYyOTYsLTk3LjI2NDU0N10sInNlbGVjdGVkR2Vvc3BhY2VJZCI6bnVsbCwiaXNFeHBsb3JhdGlvblBvcG92ZXJPcGVuIjpmYWxzZSwiYXJlU3BlZWRGaWx0ZXJzT3BlbiI6ZmFsc2V9">broadbandmapping.com</a>, Georgia Tech’s <a href="https://sites.gatech.edu/cellwatch/about/">CellWatch</a> and others integrated M-Lab data and tools into their responses to the FCC’s Broadband Data Collection initiative.</li>
</ul>
<h2 id="looking-ahead">Looking Ahead</h2>
<p>For the next year, we’ve defined the following priorities –</p>
<p>Platform</p>
<ul>
<li><strong>Increase the number of virtual sites and diversify cloud providers.</strong> To ensure that our platform represents the diversity of locations where user’s content is stored, we’ll continue pursuing partnerships with new cloud and content providers.</li>
<li><strong>Client integration partnerships.</strong> Much of M-Lab’s data is generated by people using tools created by partners external to M-Lab, or “client integrations” as we refer to them on the team. In 2023, we’re thinking about how these integrations can be better scaled and managed, and contribute to the sustainability of the M-Lab platform.</li>
<li><strong>Making it easier to contribute infrastructure to the M-Lab platform.</strong> We’re working on techniques to minimize site hosts requirements and more easily add a wider diversity of vantage points.</li>
</ul>
<p>Measurement & Data</p>
<ul>
<li><strong>Enable calibrated measurements by increasing the number of client vantage points</strong>. Building on projects such as <a href="https://www.measurementlab.net/blog/murakami/">Murakami</a> and <a href="https://internetequity.org/">Netrics</a>, our team will work to develop easy-to-use, “out-of-the-box” measurement solutions capable of measuring consistently overtime, running multiple test frameworks and differentiating between Wi-Fi and the service provider network by measuring from the access point directly.</li>
</ul>
<p>Research & Community</p>
<ul>
<li><strong>Redefine Internet Quality</strong>. Users of NDT data often consider “speed” as a proxy for the quality of an Internet connection. However, there is a growing consensus in the Internet research community that alternative metrics such as latency, jitter, and packet loss should also be considered, especially when advising policy makers and advocates. We’ll convene the M-Lab community to answer: What does it mean to have quality Internet? What should a globally recognized rubric for Internet quality include? And how can we measure it using open data and methodologies?</li>
<li><strong>Foster & Support more open Internet measurement research.</strong> Over the last year, we’ve heard lots of good ideas for research come up in conversation with community members and even our team internally. We want to find ways to help make the research happen, e.g. in partnership with sponsor organizations, supporting researchers who propose novel ways of using M-Lab data and tools in their research, sharing the research needs we’ve heard and having ourselves to prompt the community to help answer these questions.
<ul>
<li>Our first fellowship program supported by <a href="https://www.cpacket.com/">cPacket</a> will open for applications in March 2024. If you are a researcher interested in applying, watch out for an announcement on our blog.</li>
</ul>
</li>
<li><strong>More collaborations with the open Internet measurement community.</strong> 2023 brought a lot of opportunities to work with our peers in the open Internet measurement world, such as OONI, IODA, Censored Planet and more. Over the next year, we’ll work to continue this momentum by organizing more convenings to identify shared challenges, research interests and complementary solutions.</li>
<li><strong>Build out our Research and Technical Advisory Committee.</strong> Originally founded as a research consortium, M-Lab has always benefited from a generous community of experts to help guide the vision and strategy of the project. In the next year, we’ll work to scale and formalize our community of experts into more formalized committees. If you are interested in being considered, please reach out to <a href="mailto:hello@measurementlab.net">hello@measurementlab.net</a>.</li>
</ul>
<h2 id="thank-you">Thank You</h2>
<p>2023 wouldn’t be complete without thanking the folks that helped to make it happen.</p>
<ul>
<li>After a year of leadership, Katherine Townsend has moved on from her role as Director. We thank her for her contributions over 2023.</li>
<li>Thank you to our measurement, data and research partners including:
<ul>
<li>Thank you to the <a href="https://wehe.meddle.mobi/">Wehe</a> team (Dave Choffnes, Zeinab Shmeis, and Phillipa Gill), the Reverse Traceroute team (Ethan Katz-Bassett, Kevin Vermulen) for their work getting Wehe and Reverse Traceroute data, respectively, into BigQuery.</li>
<li>Thank you to Timur Friedman and the <a href="https://dioptra.io/">Dioptra</a> team for their long engagement with M-Lab and their work to publish Internet scale route trace datasets from <a href="https://iris.dioptra.io/#/">Iris</a>.</li>
</ul>
</li>
<li>Thank you to our infrastructure partners including:
<ul>
<li><a href="https://www.isc.org/">ISC</a> who has sponsored M-Lab sites for over a decade and helped provide the first minimal site configuration, which will serve as a template for more sites in the future.</li>
<li>Cogent, Hurricane Electric, Transteloco and Vodafone who continue to support infrastructure in over 20 sites globally.</li>
<li><a href="https://www.rnp.br/">RNP</a>, for working with M-Lab to bring the world’s fifth-largest digital population to NDT servers hosted in all 27 of Brazil’s federal units, with fast and reliable connectivity to major Internet exchange points (IXPs) throughout the country.</li>
</ul>
</li>
<li>On behalf of the Internet Measurement Conference Hackathon organizers (OONI, IODA, Censored Planet and Internet Society), we’d like to thank Google for arranging the conference room and lunch for the event, Google Jigsaw for the hackathon prizes, and Internet Society for the hackathon dinner.</li>
<li>OONI, IODA, Censored Planet, Internet Society and Cloudflare Radar for our ongoing collaborations.</li>
<li>Finally, we thank all the supporting partners, client integrators, analysts, developers, and users who run measurements every day around the world. Without your help, curiosity, and proof, we could not fulfill our shared mission to measure the Internet, save the data, and make it universally accessible and useful.</li>
</ul>
<h2 id="get-involved">Get Involved</h2>
<p><strong>Donate Data</strong>. As ever, testing your Internet is the easiest and best way to get involved with M-Lab — every time you measure your Internet with one of the M-Lab hosted tests, you are helping to make it possible for us to improve the Internet.</p>
<p><strong>Donate Resources</strong>. We also welcome tax-deductible donations of all sizes via our fiscal sponsor, <a href="https://donorbox.org/measurement-lab">Code for Science & Society</a>. If you have technical assets, such as cloud resources or other technical infrastructure that might benefit M-Lab, reach out and we’d love to discuss more!</p>
<p><strong>Partner with us.</strong> The M-Lab team is here to help, and can support projects that need to work with the Internet measurement data, or are interested in integrating the platform or generally thinking about ways to approach Internet measurement.</p>
<p><strong>More ways to contribute</strong>. Let’s talk about ways to partner on open Internet data, new metrics, diversifying testing infrastructure and more. If you’re interested in ways to get involved with any of the efforts mentioned above, specifically:</p>
<ul>
<li>Contributing to the M-Lab platform as a cloud or content provider</li>
<li>Sponsoring a researcher to use M-Lab data</li>
<li>Using M-Lab data in your research and letting us know</li>
<li>Joining our Research and Technical Advisory Committee</li>
<li>Be involved in our effort to redefine Internet Quality</li>
</ul>
<p>Reach out to <a href="mailto:hello@measurementlab.net">hello@measurementlab.net</a> — we’d love to chat and see what we can do next year and the years to come.</p>Lai Yi Ohlsen, Georgia Bullen2023 was a big year! Here’s a summary of what we were up to and what’s coming up next.Wehe data is now available in BigQuery2023-09-28T00:00:00+00:002023-09-28T00:00:00+00:00/blog/wehe-bigquery-announcement<p>M-Lab and the Wehe team are pleased to announce that Wehe data is available for open access in BigQuery. <!--more--></p>
<h1 id="wehe-data-is-now-in-bigquery">Wehe data is now in BigQuery</h1>
<p>Since 2020 the M-Lab platform has hosted the Wehe measurement service with a shared interest in measuring net neutrality. We are happy to announce as of September 2023 Wehe data is being automatically published to M-Lab’s public BigQuery.</p>
<h2 id="about-wehe">About Wehe</h2>
<p>The Wehe mobile application is available on iOS and Android and allows users to detect whether or not various applications are being throttled by their ISP. In other words, Wehe can tell you if your network provider is likely giving different download/upload speeds to different services (e.g., video streaming, videoconferencing, etc.)</p>
<p>Wehe uses your device to exchange Internet traffic recorded from real, popular apps like YouTube and Spotify—effectively making it look as if you are using those apps. As a result, if an Internet service provider (ISP) tries to slow down YouTube, Wehe would see the same behavior. We then send the same app’s Internet traffic, but replacing the content with something that doesn’t match any real app (we call this <em>inverted</em>), which prevents the ISPs from classifying the traffic as belonging to the app. The hypothesis is that the inverted traffic will not cause an ISP to conduct application-specific differentiation (e.g., throttling or blocking), but the original recorded traffic will. The Wehe app repeats these tests several times to rule out factors that would lead to incorrect conclusions (e.g., bad network conditions) and tell you at the end whether your ISP is giving different performance to an app’s network traffic.</p>
<p>Since 2020, the French telecommunications regulator ARCEP has <a href="https://en.arcep.fr/news/press-releases/view/n/open-internet-211220.html">integrated Wehe as their official tool</a> to verify compliance with net neutrality regulations. The Wehe team’s findings have been published in peer-reviewed conferences, covered by popular press (e.g., <a href="https://www.vice.com/en/article/j5vn9k/apple-blocking-net-neutrality-app-wehe">Motherboard</a> and <a href="https://www.bloomberg.com/news/articles/2019-08-19/wireless-carrier-throttling-of-online-video-is-pervasive-study">Bloomberg</a>) and made available to regulators such as the <a href="https://www.google.com/url?q=https://www.commerce.senate.gov/services/files/E4FB6E39-28F0-4328-902A-04F5F511825C&sa=D&source=docs&ust=1695859869686243&usg=AOvVaw035JXPubMC8oJAZQK7RAK1">FCC</a>, FTC, and State of California, as well as Canada’s CRTC.</p>
<h2 id="research-opportunities">Research Opportunities</h2>
<p>Beyond providing individual users with insight about their Internet connection, analyzing Wehe data at scale can provide insight into net neutrality violations regionally and across the world. Researchers can ask questions such as:</p>
<ul>
<li>Which ISPs in a given region are showing evidence of content-based traffic differentiation?</li>
<li>Which applications are affected by differential treatment from my ISP? Which ones are not?</li>
<li>When applications are throttled, how are they slowed down?</li>
<li>How has an ISP’s treatment of a specific application changed over time?</li>
<li>Are these network management practices in response to overloads, or instead present 24/7 and throughout the entire service area for a country?</li>
</ul>
<p>The following papers overview the Wehe measurement technique and findings of running the application in practice:</p>
<ul>
<li>Fangfan Li, Arian Niaki, David Choffnes, Phillipa Gill, Alan Mislove. <strong>A Large-Scale Analysis of Deployed Traffic Differentiation Practices</strong>. In <em>Proc. of SIGCOMM</em>, August 2019. (<a href="https://wehe.meddle.mobi/wehepaper.html">Paper</a>)</li>
<li>Arash Molavi Kakhki, Abbas Razaghpanah, Hyungjoon Koo, Anke Li, Rajeshkumar Golani, David Choffnes, Phillipa Gill, and Alan Mislove. <strong>Identifying Traffic Differentiation in Mobile Networks</strong>. In <em>Proceedings of the 15th ACM Internet Measurement Conference (IMC’15)</em>, Tokyo, Japan, October 2015. (<a href="http://david.choffnes.com/pubs/imc095-molavi-kakhkiA.pdf">Paper PDF</a>, <a href="http://david.choffnes.com/pubs/differentiation_sigcomm14_SRCposter.pdf">Poster</a>, <a href="http://david.choffnes.com/pubs/differentiation_sigcomm14_SRCtalk.pdf">Presentation slides</a>)</li>
</ul>
<p>The Wehe project has also published several other papers covering their findings and specialized measurement techniques:</p>
<ul>
<li>Z. Shmeis, M. Abdullah, P. Nikolopoulos, K. Argyraki, D. Choffnes, P. Gill **Localizing Traffic Differentiation. **To appear in Proceedings of the ACM Internet Measurement Conference (IMC), 2023. (<a href="https://nal-epfl.github.io/WeHeY/">Website</a>)</li>
<li>Fangfan Li, Abbas Razaghpanah, Arash Molavi Kakhki, Arian Akhavan Niaki, David Choffnes, Phillipa Gill, Alan Mislove. <strong>lib·erate, (n): A library for exposing (traffic-classification) rules and avoiding them efficiently</strong>. In <em>Proceedings of the 17th ACM Internet Measurement Conference (IMC’17)</em>, London, UK, November 2017. (<a href="https://david.choffnes.com/pubs/liberate-imc17.pdf">PDF</a>)</li>
<li>David Choffnes, Phillipa Gill, Alan Mislove. <strong>An Empirical Evaluation of Deployed DPI Middleboxes and Their Implications for Policymakers</strong>. In <em>Proceedings of the 45th Research Conference on Communications, Information and Internet Policy</em>, Washington, D.C., September 2017. (<a href="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2941535">PDF</a>)</li>
<li>Fangfan Li, Arash Molavi Kakhki, David Choffnes, Phillipa Gill, and Alan Mislove. <strong>Classifiers Unclassified: An Efficient Approach to Revealing IP-Traffic Classification Rules</strong>. In <em>Proceedings of the 16th ACM Internet Measurement Conference (IMC’16)</em>, Santa Monica, CA, November 2016. (<a href="https://david.choffnes.com/pubs/ClassifiersUnclassified-IMC16.pdf">PDF</a>, <a href="http://dd.meddle.mobi/classifier.html">Code and data</a>)</li>
</ul>
<p>If you’d like support using Wehe data in your research, please <a href="mailto:hello@measurementlab.net">feel free to reach out</a>.</p>
<h2 id="how-to-access-the-data">How to access the data</h2>
<p>New Wehe test data is added daily to M-Lab’s public BigQuery.</p>
<ul>
<li>To get access to all M-Lab data, including Wehe, see our <a href="https://www.measurementlab.net/data/docs/bq/quickstart/">BigQuery quick start page</a>.</li>
<li>For more sample queries and schema documentation, see the <a href="https://www.measurementlab.net/tests/wehe/#wehe-data-in-bigquery">Wehe data documentation page</a>.</li>
<li>For more details about how Wehe works, see <a href="https://wehe.meddle.mobi/td_details.html">Wehe’s Technical Documentation</a>, written for both technical and non-technical audiences.</li>
</ul>
<h2 id="community-call">Community Call</h2>
<p>To learn more about Wehe, please join us at our monthly M-Lab community call on October 5th, 2023 at 11am EDT (link to registration)[https://us02web.zoom.us/j/81069975204?pwd=VVdTOFVYTElVc1huQVVSUzJERXdndz09]. The call will be recorded for those who cannot attend.</p>
<h2 id="acknowledgement">Acknowledgement</h2>
<p>Thank you to everyone on the M-Lab and Wehe team who made this latest iteration of our collaboration possible, in particular Cristina Leon and Zeinab Shmeis for their work on making Wehe autoloadable to BigQuery!</p>Dave Choffnes, Phillipa Gill, Zeinab Shmeis, Katherine Townsend, Lai Yi OhlsenM-Lab and the Wehe team are pleased to announce that Wehe data is available for open access in BigQuery.Join us at the IMC 2023 Hackathon on Network Interference using Open Data2023-08-28T00:00:00+00:002023-08-28T00:00:00+00:00/blog/imc-hackathon-2023<p>Are you attending the <a href="https://conferences.sigcomm.org/imc/2023/">Internet Measurement Conference (IMC) 2023</a>?
<!--more--></p>
<p>Join us at the <a href="https://conferences.sigcomm.org/imc/2023/hackaton/">IMC Hackathon on Network Interference using Open Data</a> on Monday, 23rd October 2023, in Montreal, Canada. The hackathon is organized by the <a href="https://www.internetsociety.org/">Internet Society (ISOC)</a>, the <a href="https://ooni.org/">Open Observatory of Network Interference (OONI)</a>, <a href="https://www.measurementlab.net/">Measurement Lab (M-Lab)</a> and <a href="https://censoredplanet.org/">Censored Planet</a>.</p>
<p>IMC is a yearly academic conference focusing on Internet measurement and analysis. The conference is sponsored by ACM SIGCOMM.</p>
<p>This IMC Hackathon will be about exploring, analyzing, and visualizing open network measurement data with a focus on identifying Internet censorship and Internet shutdown events.</p>
<p>Participants and supporters of the hackathon include Internet measurement professionals, students, data scientists and in general anyone passionate about internet freedom. We encourage applications from people with diverse backgrounds (such as economists and social scientists) who are interested in participating in the hackathon.</p>
<p>Learn more about the hackathon <a href="https://conferences.sigcomm.org/imc/2023/hackaton/">here</a>.</p>
<p>Register to attend the hackathon <a href="https://app.smartsheet.com/b/form/8a5ab2684812497093104a31377181ac">here</a>.</p>
<p>Hackathon project ideas can be found <a href="https://docs.google.com/document/d/1bmCwU0ZJCu-xKlIBKqh3rdIJLHZ-vS7UCOfHI-0RxVE/">here</a>.</p>Lai Yi Ohlsen, Katherine TownsendAre you attending the Internet Measurement Conference (IMC) 2023?Introducing MSAK (Measurement Swiss-Army Knife)2023-07-26T00:00:00+00:002023-07-26T00:00:00+00:00/blog/introducing-msak<p>Since its inception, M-Lab’s flagship <strong>bulk transport capacity</strong> (BTC) measurement protocol, NDT, has always been using a single TCP stream. While this allowed us to get detailed diagnostic data about the user’s connection, over time many have expressed concerns about the ability of the single-stream approach to also effectively fill the link and double as a measurement of <strong>link capacity</strong>. <!--more--></p>
<p>NDT data is also frequently compared to Internet speed test services with a substantially different server placement model (usually much closer to the end user, often on the same network as the user’s ISP), which makes apples-to-apples comparisons harder. We determined that a multi-stream measurement hosted on the M-Lab platform could provide the missing data to understand the effective impact of multiple streams on throughput measurements and, at the same time, provide a platform to experiment with and shape the next generation of NDT versions.</p>
<p>In addition to this, the FCC’s <strong><a href="https://www.fcc.gov/general/measuring-mobile-broadband-performance">mobile broadband measurement program</a></strong> explicitly requires that the measurement happens over 3 concurrent streams. By supporting multi-stream measurements, we will be able to fulfill the FCC’s requirements.</p>
<p>As a result of these considerations, we are excited to introduce our newest M-Lab hosted experiment: Measurement Swiss-Army Knife, or MSAK.</p>
<h2 id="throughput-testing">Throughput testing</h2>
<p>MSAK provides a WebSocket-based throughput testing protocol with a variety of client-configurable options, including the number of streams, stream start time, congestion control algorithm used, and test duration.</p>
<p>This combination of options will allow us to test the relative performance of single and multi-stream and, by allowing different stream to go to different M-Lab servers, in many cases also allows measuring link capacity.</p>
<h2 id="latency-testing">Latency testing</h2>
<p>MSAK also provides a client-initiated UDP-based latency testing protocol, which measures latency by sending small UDP packets at a fast rate that is comparable with real-time UDP-based protocols. By running this test, the client will know its latency, jitter, and packet loss.</p>
<p>This is the first UDP-based latency test available on the M-Lab platform, and while not a complete solution, it is part of a longer-term strategy to measure more dimensions of the user’s Internet experience, including the quality of audio and video conferencing and the effects of bufferbloat on latency.</p>
<h2 id="summary">Summary</h2>
<p>MSAK is under active development, and the best way to stay up-to-date is by following its <a href="https://github.com/m-lab/msak">official GitHub repository</a>. Our first milestone is to support Georgia Tech’s CellWatch app for mobile measurement. A JavaScript client library for the throughput protocol is planned and will enable us to provide a multi-stream option on <a href="https://speed.measurementlab.net">our official speedtest</a>.</p>
<p>To learn more about the technical design, you can watch the official presentation at our community call in July (<a href="https://youtu.be/Lvo-nlX5YeM">recording</a>) (<a href="https://docs.google.com/presentation/d/1IbqMhmBwP2ul0Y7I-haIICc3tTjHsuRAKofzR5J-0gM/edit#slide=id.g255b8295bce_1_63">slides</a>)</p>Roberto D'AuriaSince its inception, M-Lab’s flagship bulk transport capacity (BTC) measurement protocol, NDT, has always been using a single TCP stream. While this allowed us to get detailed diagnostic data about the user’s connection, over time many have expressed concerns about the ability of the single-stream approach to also effectively fill the link and double as a measurement of link capacity.July 2023 Office Hours2023-07-20T00:00:00+00:002023-07-20T00:00:00+00:00/blog/july2023-officehours<p>Join us for M-Lab’s next Community Office Hours on July 20, 2023 at 11:00am EDT! <!--more--></p>
<h2 id="about">About</h2>
<p>M-Lab hosts community “office hours” on the third Thursday of every month at 11:00 AM EDT. The purpose of these calls are to:</p>
<ul>
<li>Provide hands-on guidance to community members who are new to using M-Lab data.</li>
<li>Learn about what kinds of M-Lab research users are interested in and solicit feedback about how we can make it easier.</li>
<li>Foster idea exchange and potential collaborations within the M-Lab community.</li>
</ul>
<h2 id="you-might-consider-attending-community-office-hours-if">You might consider attending Community Office Hours if…</h2>
<ul>
<li>You have a tactical question (e.g. “How do I write this query?”) or a theoretical question (e.g. “How should I shape my research question?”).</li>
<li>You are new to using M-Lab data and want more about best practices for its use.</li>
<li>You are curious abuot how others are using M-Lab data and want to share your own work as well.</li>
</ul>
<h2 id="learn-more--register">Learn More & Register</h2>
<p>You can read more about office hours <a href="https://www.measurementlab.net/learn/">here</a>, and sign-up <a href="https://docs.google.com/forms/d/e/1FAIpQLSdIBk55Jmc0lT0v0X0o-qX4t0rUrK6DZFAb0lxUU51yWwx0MQ/viewform?usp=sf_link">here</a>.</p>Lai Yi OhlsenJoin us for M-Lab’s next Community Office Hours on July 20, 2023 at 11:00am EDT!Introducing Data Transfer Limits to NDT2023-06-26T00:00:00+00:002023-06-26T00:00:00+00:00/blog/short-ndt<p>Consumer Internet speeds increasing to 1Gbps and beyond pose a scaling challenge to servers engaged in measuring global Internet quality, as every bit represents physical infrastructure and energy costs to measure that information. Maintainers of Measurement Lab are investigating options to limit the amount of data transferred by an NDT test that will reduce resource strain while maintaining quality of Internet measurement and sustain user experience.</p>
<p><!--more--></p>
<h2 id="motivation">Motivation</h2>
<p>Consumer Internet speeds increasing to 1Gbps and beyond pose a scaling challenge to servers engaged in measuring global Internet quality such as those deployed by M-Lab. For example, an NDT test will transfer 1.25 GB in 10 seconds for clients with 1 Gbps network capacity, which causes a financial strain as every bit represents both information and the physical infrastructure and energy costs to measure that information. Further, mobile speed tests may be constrained by monthly data transfer caps and could benefit from less aggressive speed test methodologies to avoid exhausting finite transfer limits; especially when their goal is to verify the availability of a certain amount of bandwidth vs. measuring the maximum throughput of a connection.</p>
<p>These observations have led maintainers of Measurement Lab to investigate options to limit the amount of data transferred by an NDT test. Specifically, we propose a client-side parameter that specifies the maximum object size to be transferred by the test. This parameter will be optional for NDT integrations. Initially, a download cap of 250MB will be applied to tests from the Internet Speed Test Google search’s NDT integration which comprises 85% of our tests and 80% of data transferred. Below we summarize the analysis that led us to choose the cap of 250 MB. We hope that this analysis will help inform others that may want to apply a cap to their own NDT integrations.</p>
<h2 id="evaluating-download-caps">Evaluating download caps</h2>
<p><img src="/images/blog/2023-06-shortndt/figure1.png" alt="Visualization of a speed test for clients with 200 Mbps, 100 Mbps and 50 Mbps.
" width="600" /></p>
<p>Figure 1: Visualization of a speed test for clients with 200 Mbps, 100 Mbps and 50 Mbps.</p>
<p>Figure 1 illustrates speed test transfers of a 125 MB object for clients with speed of 200 Mbps, 100 Mbps and 50 Mbps, respectively. The 100 Mbps client will transfer the object in exactly 10 seconds whereas the 50 Mbps client will take 20 seconds to transfer the same object. If the test is ended after 10 seconds, as is done by NDT, the 50 Mbps client will have transferred only 62.5 MB during the test. This example gives some intuition about how we may want to choose a maximum data transfer for the NDT test. Specifically, clients with speed that is lower than the speed needed to transfer the object in 10 seconds i.e. the “target rate” are not impacted by the cap because they transfer less than the cap amount in 10 seconds. However, for clients that are faster than the target rate, the test will end early e.g., after 5 seconds for a client with 200 Mbps available bandwidth.</p>
<p>The table below summarizes the maximum object sizes we consider along with their corresponding target rates. The specific set of object sizes was chosen to model a standard set of Internet speed tiers.</p>
<table>
<tr style="background-color:#ADD8E6;">
<td><strong>Size</strong>
</td>
<td>10 MB
</td>
<td>32 MB
</td>
<td>40 MB
</td>
<td>100 MB
</td>
<td>125 MB
</td>
<td>188 MB
</td>
<td>250 MB
</td>
<td>375 MB
</td>
<td>625 MB
</td>
<td>1,250 MB
</td>
</tr>
<tr>
<td><strong>Target Rate</strong>
</td>
<td>8 Mbps
</td>
<td>25 Mbps
</td>
<td>32 Mbps
</td>
<td>80 Mbps
</td>
<td>100 Mbps
</td>
<td>150 Mbps
</td>
<td>200 Mbps
</td>
<td>300 Mbps
</td>
<td>500 Mbps
</td>
<td>1 Gbps
</td>
</tr>
</table>
<h2 id="data-savings-vs-download-cap">Data savings vs. download cap</h2>
<p><img src="/images/blog/2023-06-shortndt/figure2.png" alt=" Percentage data savings vs. download cap. Percentage savings is computed over all tests in Q1 2023." width="600" /></p>
<p>Figure 2. Percentage data savings vs. download cap. Percentage savings is computed over all tests in Q1 2023.</p>
<p>Figure 2 shows the percent of data transferred by NDT tests that would be saved if all tests were capped at a given value. Data savings start at 91% for a cap of 10 MB per test and rapidly drop off with a savings of only 5% for a 625 MB cap. We note that 100-250 MB seem to provide a decent middle ground with data savings of 25-50% for these caps.</p>
<h2 id="accuracy-vs-download-cap">Accuracy vs. download cap</h2>
<p>We next study the accuracy of NDT tests with different download caps. For this analysis, we analyze traces of existing NDT tests and emulate the bandwidth estimation process if the test had stopped after a fixed amount of data was transferred. We compare this estimated throughput with the throughput that was returned by the full test to study the accuracy of NDT with different sized download caps.</p>
<p><img src="/images/blog/2023-06-shortndt/figure3.png" alt="Percentage error vs. download cap showing the median-95th percentile error for all tests with a given download cap." width="600" /></p>
<p>Figure 3. Percentage error vs. download cap showing the median-95th percentile error for all tests with a given download cap.</p>
<p>Figure 3 shows the accuracy of NDT tests for the different size caps. Overall accuracy is quite good, even for smaller object sizes with the median error near zero with an object size of 100 MB and the 75th percentile reaching zero at 125MB. The 90th and 95th percentile errors take longer to drop off with an object size of 250 MB having a 95th percentile error of 5% across all tests. 2</p>
<p>This result may seem to imply that we can use a very small object size (e.g., 100 MB) while maintaining relatively good accuracy, especially if we focus on the median and 75th percentile error. However, as we noted above, accuracy is dependent on the client speed as well as the object size.</p>
<p><img src="/images/blog/2023-06-shortndt/figure4.png" alt="95th percentile error vs. download cap for different client speeds." width="600" /></p>
<p>Figure 4. 95th percentile error vs. download cap for different client speeds.</p>
<p>Figure 4 illustrates the relationship between the download cap and 95th percentile error for different client speeds. While the 95th percentile accuracy is fairly good for client speeds up to 250 Mbps with less than 5% error, the error increases with 95th percentile error of 18-30% for client speeds above 250 Mbps.</p>
<p><img src="/images/blog/2023-06-shortndt/figure5.png" alt="Distribution of client speeds for NDT tests." width="600" /></p>
<p>Figure 5. Distribution of client speeds for NDT tests.</p>
<p>As we’ve seen, our overall accuracy is high even for small object sizes, but the accuracy is very dependent on the client’s speed. Figure 5 shows the breakdown of client speeds for NDT tests. We find that 91% of the clients have speeds of <=250Mbps. From this we can conclude that the overall accuracy of the shorter tests that we observe is related to the NDT client population.</p>
<h2 id="choosing-a-cap">Choosing a cap</h2>
<p>Based on our analysis there are a few options for a download cap. We could pick a small value like 100 MB which gives 50% data savings and a median accuracy of close to 0 or we can go with a larger value that gives higher accuracy.</p>
<p>In the end, we have decided on a cap of 250 MB for the Internet speed test. This corresponds to a client speed of 200 Mbps if the object is transferred in exactly 10 seconds. 200 Mbps is also substantially above the currently defined definition for broadband by the FCC of 25 Mbps for downloads and 3 Mbps for uploads. This cap will yield 25% data savings with an overall 95th percentile error of 5%.</p>
<h2 id="community-call">Community call</h2>
<p>Measurement Lab held a call with community members on June 1st, 2023 to announce these changes to NDT. For more information, the recording can be found <a href="https://youtu.be/J6ueXt98WKE">here</a> and the slides <a href="https://docs.google.com/presentation/d/1Ls3ivzf3Ya-7WL7w4JkK_MyVxkf4PDsgbh47oKI3-Oc/edit#slide=id.g8dc385d038_0_802">here</a>.</p>
<h2 id="anticipating-concerns">Anticipating concerns</h2>
<p>While Measurement Lab’s intention is to assess the throughput of a connection, we recognize that most contributors running a test are keen to understand whether the Internet service they have paid for is meeting its advertised benefits. We recognize that many companies and households pay for speeds beyond 250 Mbps, reaching upwards of 1Gig and higher, and that shorter NDT Tests may be inadequate for the 9% of contributors seeking to test their bandwidth. We note that the object size limit is strictly opt-in at the client side and higher bandwidth integrations can still leverage an unlimited NDT test. We do stress that the need of the majority of Internet users maxes out well below 250 Mbps and are curious about the perception that more and faster is better, and whether we can improve our understanding of quality Internet over quantity.</p>
<h2 id="call-to-action">Call to action</h2>
<p>While we have focused on limiting data transferred by the speed test as one way to allow our tests to scale in the face of higher consumer broadband speeds, there is an opportunity for more research in this space. We are open to collaborating with those exploring novel research for speed tests that are able to measure higher speeds, with high accuracy while conserving resources. Please reach out if you are working toward improving this space!</p>Phillipa Gill, Cristina Leon, Katherine TownsendConsumer Internet speeds increasing to 1Gbps and beyond pose a scaling challenge to servers engaged in measuring global Internet quality, as every bit represents physical infrastructure and energy costs to measure that information. Maintainers of Measurement Lab are investigating options to limit the amount of data transferred by an NDT test that will reduce resource strain while maintaining quality of Internet measurement and sustain user experience.Introducing M-Lab Community Office Hours2023-06-16T00:00:00+00:002023-06-16T00:00:00+00:00/blog/introducing-office-hours<p>Are you new to M-Lab data? Do you have questions about how to use it or how it can fit into your research? <!--more--></p>
<h2 id="about">About</h2>
<p>Starting June 22, 2023 M-Lab will begin to host community “office hours” on the third Thursday of every month at 11:00 AM EDT. The purpose of these calls are to:</p>
<ul>
<li>Provide hands-on guidance to community members who are new to using M-Lab data.</li>
<li>Learn about what kinds of M-Lab research users are interested in and solicit feedback about how we can make it easier.</li>
<li>Foster idea exchange and potential collaborations within the M-Lab community.</li>
</ul>
<h2 id="notes">Notes</h2>
<ul>
<li>Both tactical (e.g. “How do I write this query?”) and theoretical (e.g. “How should I shape my research question?”) questions are welcome.</li>
<li>As we begin these office hours <strong>we are limiting each session to five participants</strong> to ensure focused time for each person that joins. If more people are interested than spots available, we’ll add more sessions!</li>
<li>While these sessions will prioritize community members who are just getting started with M-Lab data, more advanced users are welcome to express interest via the sign-up form (just mention it in the notes). If there is enough interest we’ll consider adding other formats.</li>
<li>If you are interested, but the time does not work for you, fill out the sign-up form and let us know in the notes.</li>
</ul>
<h2 id="sign-up">Sign Up</h2>
<p>To sign up, fill out our <a href="https://docs.google.com/forms/d/e/1FAIpQLSdIBk55Jmc0lT0v0X0o-qX4t0rUrK6DZFAb0lxUU51yWwx0MQ/viewform?usp=sf_link">sign-up form</a>.</p>Lai Yi OhlsenAre you new to M-Lab data? Do you have questions about how to use it or how it can fit into your research?Ending support for ndt5+raw protocol and mlab-ns2023-05-08T00:00:00+00:002023-05-08T00:00:00+00:00/blog/retiring-ndt5-raw<p>After January 2024, M-Lab will no longer support the legacy ndt5+raw protocol or
mlab-ns.appspot.com service. There are alternatives available for migration now.
<!--more--></p>
<h2 id="summary">Summary</h2>
<p>For the last three years, ndt7 and the Locate v2 API have served the vast
majority of measurements to M-Lab. And, throughout that time, we have maintained
support for the ndt5+raw protocol and the mlab-ns.appspot.com location service
for backward compatibility with legacy clients. After January 2024, that support
will end.</p>
<p>We recommend that you begin transitioning to ndt7 and the Locate v2 API as soon
as possible to ensure that you have enough time to migrate existing clients and
avoid any disruption.</p>
<p>We are deeply grateful for the time and effort already taken to integrate with
the M-Lab platform. And, we understand that this may be disruptive for you and
your users and apologize for any inconvenience this may cause. We will continue
to provide support and assistance during the migration period.</p>
<p>If you have any questions or concerns, please let us know.</p>
<h2 id="why-is-this-necessary">Why is this necessary?</h2>
<p>We created ndt5 as part of the <a href="/blog/the-platform-has-landed/">platform migration</a> to preserve
backward compatibility for clients dating back to the beginning of M-Lab and the
web100 platform, 2009 to 2019. In 2020, we <a href="/blog/ndt7-introduction">introduced ndt7</a> with the
<a href="/develop/locate-v2/">Locate v2 API</a> as a successor to ndt5, which addressed many
limitations; ndt7 is a simple protocol (single connection), uses standard ports
(80, 443), uses modern congestion control (BBR), and supports admission control
natively.</p>
<p>For ndt7 clients, we <a href="/blog/ndt7-access-tokens">required admission control</a>. The ndt5+ws and
ndt5+wss protocols also support admission control but are not yet required.
Unfortunately, the ndt5+raw protocol (since it was developed before modern web
standards) cannot support admission control in a backward compatible way. The
only remaining clients of mlab-ns.appspot.com are those using ndt5+raw.</p>
<p>Finally, since M-Lab added <a href="/blog/2022-mlab-to-the-cloud/">virtual servers</a> from <a href="/blog/virtual-sites-gcp/">Google Cloud</a> to
the platform, and because Cloud servers operate with a different cost structure
for egress traffic, we must enforce admission control for all clients to these
servers. So, virtual servers have already disabled the ndt5+raw protocol and
require access tokens for ndt5+ws and ndt5+wss. By retiring the ndt5+raw
protocol, we can require admission control for all clients on all server types,
and retire mlab-ns.appspot.com.</p>
<h2 id="migration-to-ndt7ws-and-ndt7wss">Migration to ndt7+ws and ndt7+wss</h2>
<p>Today, M-Lab supports multiple official and unofficial <a href="https://github.com/m-lab/ndt-server#clients">ndt7 clients</a>.</p>
<ul>
<li>https://github.com/m-lab/ndt7-js (official, javascript)</li>
<li>https://github.com/m-lab/ndt7-client-go (official, golang)</li>
<li>https://github.com/m-lab/ndt7-client-ios (unofficial, swift)</li>
<li>https://github.com/m-lab/ndt7-client-android (unofficial, kotlin)</li>
<li>https://github.com/m-lab/ndt7-client-android-java (unofficial, java)</li>
</ul>
<p>All of these clients use the Locate v2 API, so support admission control
natively and require no customizations.</p>
<h2 id="migration-to-locate-v2-api">Migration to Locate v2 API</h2>
<p>If you have developed a custom client, consider sharing your work publicly so
others can benefit from a reusable and verifiable open source test client. We
would be happy to add a link to help others find you work.</p>
<p>If your custom client depends on mlab-ns.appspot.com, you can find guidance for
<a href="/develop/locate-v2/">migrating to the Locate v2 API here</a>.</p>
<h2 id="let-us-know">Let us know</h2>
<p>If you have exceptional concerns about the retirement of ndt5+raw or
mlab-ns.appspot.com, <a href="mailto:support@measurementlab.net">please let us know</a>.</p>Stephen SolteszAfter January 2024, M-Lab will no longer support the legacy ndt5+raw protocol or mlab-ns.appspot.com service. There are alternatives available for migration now.