TechPrufer
Get started
Meta
Meta logoMeta · Feature trail

Internet Connectivity Expansion and Infrastructure

82 posts · Mar 2012 – Apr 2026

What this trail covers

Meta's efforts to expand internet access globally have evolved significantly, encompassing the establishment of the Connectivity Lab to develop novel aerospace and communication technologies like planes and satellites for internet delivery. This initiative is part of the broader Internet.org mission to improve and extend internet access to underserved regions. The development of advanced connectivity solutions, including the first full-scale model of Aquila (a high-altitude, long-endurance aircraft), represents a key milestone in this ongoing effort.

All Meta trails
  1. Escaping the Fork: How Meta Modernized WebRTC Across 50+ Use CasesThis post details Meta's strategy for modernizing its WebRTC implementation, which powers real-time audio and video across numerous services. It addresses the challenge of maintaining a forked open-source project within a monorepo by introducing a dual-stack architecture and a shim layer. This allows for safe A/B testing of new upstream versions against legacy versions, enabling continuous upgrades and improved performance, binary size, and security across over 50 use cases. The post also outlines technical solutions for symbol collision resolution, backward compatibility, runtime dispatch, and automated shim generation.
  2. Announcing the Completion of the Core 2Africa System: Building the Future of Connectivity TogetherThis post announces the completion of the core 2Africa infrastructure, the world's longest open access subsea cable system. It details the unprecedented scale and reach, connecting 3 continents and 33 countries, enabling connectivity for 3 billion people. The post highlights the engineering innovations, including the deployment of 16-fiber-pair SDM technology (double the capacity of older systems), undersea optical wavelength switching, increased burial depth, dual trunk powering architectures, and branching unit switching for reliability. It also describes the logistical challenges, involving 35 offshore vessels and extensive shore-end operations. The economic impact is quantified, with an expected contribution of $36.9 billion to Africa's GDP within 2-3 years, driven by increased bandwidth, lower wholesale prices, market competition, and support for modern digital infrastructure like cloud services and 5G.
  3. 10X Backbone: How Meta Is Scaling Backbone Connectivity for AIThis post details Meta's "10X Backbone" initiative, focusing on scaling the Express Backbone (EBB) network to support the increasing demands of AI workloads. It introduces three key techniques: DC Metro Architecture for simplified and scalable DC connectivity, IP Platform Scaling (both scaling up with larger chassis/faster interfaces and scaling out by adding planes/devices), and IP/Optical Integration using ZR technology for significant power and space efficiency. The post also outlines solutions for AI cluster expansion, including FR, LR, and ZR plugs with DWDM, to address varying reach requirements.
  4. Introducing the Candle Subsea Cable, Updates to Our Asia-Pacific Connectivity ProjectsThis post announces the introduction of the Candle submarine cable system, which will connect Japan, Taiwan, the Philippines, Indonesia, Malaysia, and Singapore with 570 terabits per second (Tbps) of capacity. It also provides updates on the Bifrost cable system, which now connects Singapore, Indonesia, the Philippines, and the United States, with Mexico expected in 2026, adding over 260 Tbps of redundancy. The Echo cable now delivers 260 Tbps of capacity between Guam and California, and the Apricot system is available between Japan, Taiwan, and Guam, with future extensions to the Philippines, Indonesia, and Singapore, providing 290 Tbps of capacity. Together, these cables aim to enhance intra-Asia connectivity and transpacific bridges to the Americas, complementing other investments like 2Africa and Project Waterworth.
  5. Extending the Malbec subsea cable to Southern BrazilThis post details the extension of the Malbec subsea cable to Porto Alegre, Brazil, in partnership with V.tal. This expansion will bring over 84 terabits of international capacity to Southern Brazil, improving digital infrastructure scale and reliability, and positioning Porto Alegre as a digital hub. It also highlights the activation of a Point of Presence (PoP) in Porto Alegre for efficient local content delivery.
  7. Taking the plunge: The engineering journey of building a subsea cableThis post details the engineering journey of building subsea cables, highlighting their critical role in Meta's global infrastructure for transporting information and content. It discusses the innovation and quality prioritized in designing and deploying these cables, referencing Project Waterworth as a recent example of large-scale subsea cable projects.
  8. Unlocking global AI potential with next-generation subsea infrastructureThis post announces Project Waterworth, Meta's most ambitious subsea cable endeavor to date. It will span over 50,000 km, connecting five major continents and utilizing the highest-capacity technology available, including 24 fiber pairs. The project emphasizes enhanced cable resilience through advanced engineering design, first-of-its-kind routing to maximize cable laid in deep water (up to 7,000 meters), and enhanced burial techniques in high-risk areas to prevent damage. The stated goal is to provide the abundant, high-speed connectivity needed to drive AI innovation globally.
  9. Threads has entered the fediverseThis post details the technical integration of Threads with the fediverse using the ActivityPub protocol. It explains the challenges and solutions for federating content, specifically addressing the handling of quote posts by implementing FEP-e232 and Misskey's `_misskey_quote` property. It also discusses the complexities of federated and non-federated interactions within Threads and outlines a phased approach for future fediverse features.
  10. Meta’s Evenstar is transitioning to OCP to accelerate open RAN adoptionMeta is transferring its IP for Evenstar, a program to accelerate the adoption of open RAN technologies, to the Open Compute Project (OCP). This includes contributing Evenstar's radio unit design, firmware, design files, and test scripts to OCP, making it the industry's first open, white box radio unit solution. The design features an advanced 4T4R MIMO transceiver, a modular architecture for mixing and matching components from different suppliers, and relies on off-the-shelf discrete components for lower BOM cost and independent control of power amplifier modules. The firmware handles complex signal processing algorithms like digital pre-distortion (DPD) and crest factor reduction (CFR). This contribution aims to accelerate open RAN commercialization by enabling the ecosystem to build modularized and efficient radio unit platforms.
  11. Bombyx is being licensed for product developmentThis post details the licensing of Meta's Bombyx aerial fiber deployment robot technology to Hibot. Bombyx is a robot designed to weave fiber-optic cables over powerlines, addressing the cost and complexity of traditional fiber installation. Key innovations include a miniature fiber cable, spool-free geometry, and the robot's ability to traverse energized conductors and cross obstacles using machine vision and custom sensors. The licensing agreement will enable Hibot to further develop the technology for product deployment, infrastructure inspection, and grid modernization.
  12. A new way to survey potential fiber routes — without access to paved roadsThis post introduces a novel technique for optical fiber route surveys in areas with limited paved roads, specifically in the Democratic Republic of the Congo. The method combines dynamic cone penetrometers (DCP) with gamma-ray spectrometers mounted on a high-clearance 4x4 vehicle. DCP provides localized soil density data, while spectrometers offer broader soil composition and heterogeneity analysis. By using a small set of DCP tests to calibrate a larger volume of spectrometer data, the process minimizes the need for extensive on-the-ground testing. This approach, integrated with mobile GPS data for obstacle and sensitive area identification, allows for detailed geographical information system maps with soil condition data without extending survey timelines. This innovation aims to reduce cost risk and improve the financial viability of fiber projects in emerging markets.
  13. Malbec subsea cable connects Argentina and Brazil with the rest of the worldLaunched Malbec, a 2,500 km subsea cable connecting Argentina and Brazil, doubling previous capacity to Argentina and increasing internet penetration. Utilized spatial division multiplexing and advanced fiber optic technology in repeaters and transponders for improved capacity and reliability. This initiative is part of Meta's broader commitment to building scalable infrastructure through open, flexible architecture partnerships.
  14. 2Africa Pearls subsea cable connects Africa, Europe, and Asia to bring affordable, high-speed internet to 3 billion peopleThis post announces the expansion of the 2Africa subsea cable system with the addition of the 2Africa Pearls segment. This extension significantly increases the total length of the cable system to over 45,000 kilometers, making it the longest ever deployed. The Pearls branch adds new landing locations in Oman, UAE, Qatar, Bahrain, Kuwait, Iraq, Pakistan, India, and Saudi Arabia. The combined 2Africa system will now provide connectivity to 3 billion people, a substantial increase from the initial goal of 1.2 billion, and will interconnect 33 countries across Africa, the Middle East, and Europe. This investment aims to bring affordable, high-speed internet to a larger population, particularly in Africa, which is currently the least connected continent.
  15. Apricot subsea cable will boost internet capacity, speeds in the Asia-Pacific regionAnnounces participation in the Apricot subsea cable system, a 12,000-kilometer cable connecting Japan, Taiwan, Guam, the Philippines, Indonesia, and Singapore. The cable will have an initial design capacity of over 190 terabits per second and is expected to launch in 2024. This project is part of Meta's ongoing effort to expand global network infrastructure and improve internet access in the Asia-Pacific region.
  16. Driving towards an open internet ecosystem to help tackle the digital divideThis post details Meta's ongoing efforts and partnerships to bridge the digital divide by expanding internet connectivity. Key contributions include: announcing a partnership with Liquid Intelligent Technologies to build 2,200 km of fiber in the Democratic Republic of Congo, expanding terrestrial fiber investments in Pakistan and Indonesia, announcing the integration of Magma with AWS for faster network deployments, highlighting the growth of the Evenstar program for OpenRAN acceleration, and detailing new analytics and monetization capabilities for the Express Wi-Fi platform. It also mentions the commercial launch of MikroTik's 60 GHz product line based on Terragraph technology.
  17. Running Border Gateway Protocol in large-scale data centersThis post details the scalable design, software implementation, and operations of Facebook's data center routing design based on Border Gateway Protocol (BGP). It highlights how BGP can form a robust routing foundation within data centers when tightly integrated with topology, configuration, switch software, and operational pipelines. The post introduces a uniform AS numbering scheme, hierarchical route summarization, and an in-house BGP agent with a robust testing and deployment framework to ensure fast, incremental updates and high availability for services.
  18. Advancing connectivity between the Asia-Pacific region and North AmericaThis post announces the development of two new subsea cables, Echo and Bifrost, connecting Singapore and Indonesia to North America. These cables will increase transpacific capacity by 70% and provide a new, diverse route across the Java Sea. This investment complements existing infrastructure and aims to support the growing demand for internet access in the Asia-Pacific region, contributing to economic growth and digital inclusion.
  19. 2020 year in review: Connectivity innovations, faster apps, and progress toward net zeroThis post details several connectivity innovations from 2020. It highlights the development of an aerial fiber deployment solution using autonomous robots, the 2Africa subsea cable project aimed at connecting Africa and the Middle East, and the SuperCell solution for rural communities. It also introduces the Evenstar program for affordable broadband through OpenRAN, and mentions the continued adoption of Magma, an open mobile core network platform. Additionally, it touches on the optimization of hardware for microservices with SoftSKU and Accelerometer, and the use of QUIC and HTTP/3 for internet traffic.
  20. Evenstar: The road to an alternative infra supply chain for RANThis post details the progress of the Evenstar program, a collaboration focused on building general-purpose Radio Access Network (RAN) reference designs for 4G and 5G networks within the Open RAN ecosystem. It announces the completion of lab validation for the first Evenstar remote radio unit (RRU) product and the commencement of engineering work on the distributed unit (DU). The program aims to decouple RRU, DU, and control unit (CU) software, enabling mobile network operators to select best-of-breed components from various vendors, fostering innovation and competition to reduce total cost of ownership. The post highlights the collaboration with industry partners and the alignment with 3GPP and O-RAN specifications, with a roadmap for second-generation RRUs and DU hardware development.
  21. SuperCell: Reaching new heights for wider connectivityThis post details the development and testing of SuperCell, a wide-area coverage solution for increasing mobile connectivity in rural communities. It introduces SuperCell as a cost-effective alternative to traditional macrocell sites, leveraging tall towers (up to 250 meters) and high-gain, narrow-sectored antennas to extend mobile data coverage range and capacity. The post quantifies the coverage advantage through propagation testing and drive measurements, analyzes angular spread to address intersector interference and capacity scaling, and presents LTE field test results demonstrating throughput and voice/video call capabilities at long ranges. It also compares SuperCell to other connectivity solutions and provides a playbook for the telecom industry.
  22. The economic impact of subsea cables in AfricaThis post quantifies the economic impact of subsea cables in Africa, specifically highlighting the 2Africa project. It presents research from RTI International estimating increases in African GDP (0.42-0.58%) and employment within the first 2-3 years of the 2Africa cable going live. It also details country-specific impacts on employment and GDP per capita for Nigeria, Democratic Republic of Congo, Kenya, South Africa, Mozambique, and Tanzania, based on historical subsea cable landings. The post emphasizes that subsea cables and broadband infrastructure investments are among the most effective development policies for economic growth in Africa.
  23. Making aerial fiber deployment faster and more efficientThis post details the development of an aerial fiber deployment system that utilizes a robot to attach specialized fiber-optic cable to medium-voltage (MV) power lines. The system addresses challenges in cost-effective fiber deployment by leveraging existing electrical infrastructure, reducing the need for extensive make-ready work. Key innovations include a robot designed for safe operation on energized lines, a fiber cable design optimized for size and weight (24-strand cable for ~24,000 homes), and a helical wrap method adapted for MV conductors. The system aims to achieve a 3x to 5x cost reduction in aerial fiber construction and enable fiber deployment closer to a larger portion of the global population.
  24. Building a transformative subsea cable to better connect AfricaThis post announces the 2Africa subsea cable project, a partnership to build the most comprehensive subsea cable serving the African continent and Middle East. It highlights the cable's length (37,000 km), capacity (nearly three times current subsea cables serving Africa), and its role in providing internet capacity, redundancy, and reliability. It also details technological advancements used, including aluminum conductors, Spatial Division Multiplexing (SDM1) for up to 16 fiber pairs, and Wavelength Selective Switching (WSS)-ROADM for flexible capacity management. The design emphasizes resiliency with increased burial depth and optimized routing. This project is presented as a continuation of Meta's efforts to expand global network infrastructure and foster an open and inclusive internet ecosystem.
  25. Implementing a secure web-based proxy service for DiscoverThis post details the technical challenges and architectural solutions for building a secure web-based proxy service (Discover) to provide free internet access. It describes the evolution from Free Basics, addressing limitations of previous approaches like subdomain allocation and IPv4-in-IPv6 encapsulation. The core innovation is an origin-collapsed domain namespace under discoverapp.com, allowing operators to easily allowlist traffic to a single domain. Key technical details include URL rewriting to encode third-party origins, server-side cookie management encrypted with a client-side key (ick), and mechanisms to prevent JavaScript-based cookie fixation and session hijacking using an outer trusted frame and localStorage for storing a derived cookie integrity token (ickt).
  26. Accelerating innovations in infrastructure and advancing global connectivity with our partnersThis post details several new initiatives and partnerships aimed at expanding global internet connectivity. It announces new collaborations for Terragraph to bring affordable, high-speed broadband to urban, suburban, and rural communities, including commercial deployments with Common Networks and Cambium Networks, and pilot programs in Puerto Rico and Malaysia. The launch of Evenstar remote radio units (RRUs) is announced to accelerate OpenRAN adoption in 4G/5G networks. Updates on Magma open-source software highlight its adoption for fixed wireless access and offloading cellular data at the network edge, with successful interoperability tests with RAN vendors. Progress with satellite operators (Viasat, Hughes Network Systems, Hispasat, YahClick, Thaicom) and network-as-a-service (NaaS) providers like Internet para Todos Peru (IpT) is reported to strengthen rural connectivity. New fiber builds in the Democratic Republic of the Congo (DRC) and Indonesia, and expansion in Nigeria are detailed, along with work on reducing the cost of aerial fiber deployments. The Express Wi-Fi platform is expanding in Indonesia with BaliTower and welcoming new technology partners (Cisco, Edgecore, IgniteNet, Inventum, Tanaza).
  27. Groundbreaking event for our newest long-haul fiber routeThis post announces a groundbreaking event for a new long-haul fiber route connecting data centers in Ohio, Virginia, and North Carolina. This initiative is part of Meta's ongoing network infrastructure build to provide resilient and scalable paths to data centers, ensuring fast, efficient, and reliable access for users. The new fiber route crosses West Virginia and is a component of connecting 11 U.S. data center locations via redundant, long-haul networks.
  28. MaRS: How Facebook keeps maps current and accurateThis post introduces MaRS (Machine-Augmented Automatic Review System) and LoChas (Logical Changesets) as new tools developed to efficiently and accurately ingest and maintain map data from OpenStreetMap. These tools address the challenges of keeping local map copies current and accurate by breaking down changesets into individual operations and using machine learning for automated review, thereby reducing the workload for human intervention and ensuring geometric correctness and idempotency of updates.
  29. Extending DHCPLB: The path from load balancer to serverThis post details the evolution of Meta's DHCP infrastructure, moving from ISC-DHCP to Kea, and finally to an extended DHCPLB that acts as a server. It highlights the challenges of single-threaded servers like Kea leading to performance bottlenecks and packet drops, and how the Go-based DHCPLB with its multithread design and the new DHCP library addresses these issues, resulting in significantly improved throughput and reduced server count. The post also emphasizes the open-sourcing of the DHCPLB framework and the underlying DHCP library.
  30. Using aluminum to overcome capacity challenges in submarine cable systemsIntroduced aluminum as a conductor for submarine cable systems as an alternative to copper. This development allows for a lower cable voltage drop, enabling a higher number of fiber pairs per cable, which is a significant step towards increasing overall cable capacity and enabling space-division multiplexing systems. This is the first major development in submarine cable technology since the early 1990s.
  31. Building backbone network infrastructureThis post details Meta's continued investment in backbone network infrastructure, specifically focusing on building new long-haul fiber routes to connect data centers in Ohio, Virginia, and North Carolina. It also highlights past investments in terrestrial fiber (e.g., New Mexico to Texas) and subsea fiber-optic cables (Marea, Havfrue, Jupiter). The company's strategy involves partnering on existing fiber, investing in new infrastructure, and leading investments in new fiber routes for resiliency and scale. A key aspect is making excess capacity on these fiber routes available to third parties, operating as a wholesale provider through a subsidiary.
  32. Enabling better global connectivity through new partnerships and technologiesThis post announces new partnerships and technologies aimed at expanding global internet connectivity, particularly in rural areas. Key initiatives include: 1. 'Internet para Todos' (IpT) Peru, a partnership with Telefonica, IDB Invest, and CAF to create an open-access wholesale rural mobile infrastructure operator. 2. Collaboration with Viasat to accelerate the deployment of satellite-enabled Community Wi-Fi hotspots in Mexico, leveraging open population maps for optimal deployment locations. 3. The announcement of Magma, an open-source software platform for simplifying and extending mobile network deployments, with initial partners Telefonica and BRCK. 4. Investment in MainOne for a 750 km open-access fiber deployment in Nigeria, building on previous fiber projects in South Africa and Uganda. 5. Expansion of the Express Wi-Fi program with new partners (CellC, Vodafone, Globe) and technology partners (Nokia, Aruba, Edgecore Networks), integrating SOMA technology into Cambium Networks' Wi-Fi products. 6. Growth of the Terragraph ecosystem with new partners (MikroTik, Cambium Networks) for millimeter-wave gigabit wireless connectivity in urban and suburban communities. The post also highlights participation in the Telecom Infra Project (TIP) and updates shared at Mobile World Congress 2019.
  33. Open-sourcing Magma to extend mobile networksThis post announces the open-sourcing of Magma, a software platform designed to help operators deploy mobile networks quickly and easily. Magma provides a software-centric distributed mobile packet core and tools for automating network management. It is intended to extend existing network topologies to rural deployments, private LTE networks, or wireless enterprise deployments, enabling new network archetypes with fast, incremental upgrade cycles and continuous software integration. Magma simplifies network operations through automation of element configuration, software updates, and device provisioning, making LTE deployment as simple as WiFi access point deployment. It works with existing LTE base stations and can federate with traditional mobile cores.
  34. A new predictive model for more accurate electrical grid mappingThis post details the development of a new predictive model for mapping medium-voltage (MV) electrical infrastructure using publicly available data sets. The model leverages nighttime radiance data from VIIRS day-night band sensors, combined with a modified Dijkstra's shortest path algorithm, to infer grid paths. The output is released for six countries and includes detailed documentation and code for replication. The model achieves approximately 70% accuracy within one kilometer of known grid locations based on public data alone, with potential for further improvement.
  35. TIP Summit 2018: Advancing efforts to improve global connectivityThis post details Facebook's ongoing contributions to the Telecom Infra Project (TIP) and its efforts to improve global connectivity. It highlights the expansion of the Terragraph ecosystem with new trials and chipset support, advancements in rural deployments in Latin America and Africa using technologies like OpenRAN and OpenCellular, and the promotion of new network planning tools that leverage connectivity analytics. Key takeaways from early deployments of Terragraph (e.g., 1 Gbps speeds, 99.5% availability) and rural connectivity solutions are shared, emphasizing the financial viability and accelerated deployment potential of these technologies. The post also announces next steps in working with the TIP community to build and deploy technologies for extending networks and upgrading speeds.
  36. FBOSS: Building switch software at scale and in the openThis post details the development and open-sourcing of FBOSS (Facebook open switching system), a custom switch software designed for large-scale, rapidly growing networks. It highlights FBOSS's leaner architecture compared to standard switch software, its modular abstractions, and its role in enabling a 30x increase in FBOSS deployments over two years. The post also mentions FBOSS's importance in implementing IPv6 and the Fabric Aggregator for inter-building connectivity.
  37. Express Wi-Fi Technology Partner program: Getting more people online with fast and affordable Wi-FiIntroduced the Express Wi-Fi Technology Partner program, allowing access point manufacturers to build hardware compatible with Express Wi-Fi. This program aims to improve Wi-Fi access point capabilities by enabling more local processing for authentication and accounting, enhancing device detection of registration pages, and improving data accounting accuracy. New access points can also account for different traffic classes, allowing for tiered service offerings. This initiative aims to make it easier for operators and entrepreneurs to deploy and manage hotspots, ultimately connecting more people with fast and affordable Wi-Fi.
  38. Open-sourcing Facebook’s Solar-Powered Aircraft Design ToolsThis post details the open-sourcing of FBHALE (Facebook High Altitude Long Endurance), a multidisciplinary optimization (MDO) design framework for solar-powered aircraft. It explains the need for MDO due to trade-offs between aerodynamic performance and structural mass, outlines the framework's logic including aerodynamic and structural modeling, and presents example applications comparing single-boom and dual-boom configurations. The post highlights the potential for future developments in control systems and power systems, and its use in airworthiness certification.
  39. Announcing tools to help partners improve connectivityThis post introduces two new tools, Advanced Network Planning and Actionable Insights, to help connectivity partners improve network infrastructure. Advanced Network Planning uses satellite imagery, computer vision, and machine learning to identify underserved areas and optimize network design for cost and performance. Actionable Insights, with its Network Insights and Market Insights components, leverages aggregated and de-identified data from Facebook users (with location services enabled) to provide network performance maps, identify congestion, and offer market share and device adoption insights to connectivity providers. The tools are designed with privacy by design principles, sharing only aggregated and de-identified information.
  40. High altitude connectivity: The next chapterThis post announces a strategic shift in Meta's approach to High Altitude Platform Station (HAPS) connectivity. Instead of continuing to design and build its own aircraft (Aquila), Meta will cease operations at its Bridgwater facility and focus on partnering with companies like Airbus. The company will continue to contribute to other critical system components, such as flight control computers and high-density batteries, and will actively engage in policy efforts to secure more spectrum for HAPS and influence aviation regulations. This marks a transition from direct aircraft development to a collaborative model focused on system integration and policy advocacy.
  41. How IPv6 deployment is growing in U.S. and other countriesThis post details the growth of IPv6 deployment across the US and other countries, highlighting its performance benefits over IPv4 and the critical need for migration due to IPv4 address exhaustion. It shares data on Facebook's IPv6 traffic, including the impact of adjusting the Happy Eyeballs algorithm timeout to improve application responsiveness. The post also notes the significant IPv6 adoption by US mobile carriers and ISPs, as well as exponential growth in countries like Taiwan, India, Mexico, and Vietnam, encouraging further industry-wide adoption.
  42. Scaling the Facebook backbone through Zero Touch ProvisioningThis post introduces and details Facebook's Zero Touch Provisioning (ZTP) system, named Vending Machine, for automating the provisioning of its global backbone networks. It describes the challenges of scaling heterogeneous IP and optical networks, the limitations of previous manual and console-based approaches, and the development of Vending Machine. Key technical contributions include the use of Python agents executed on network devices via ZTP to download instructions, execute commands, and report logs, and a flexible workflow framework composed of independent 'steps' that can be developed and deployed iteratively. The system aims to replace manual Method of Procedures (MOPs) with a single command for configuring any network device.
  43. Fabric Aggregator: A flexible solution to our traffic demandThis post introduces the Fabric Aggregator, a distributed network system designed to scale the network capacity between data center fabrics. It addresses the growing demand for bandwidth driven by immersive experiences like live video, 360 photos/videos, and VR. The Fabric Aggregator utilizes a disaggregated approach with simple, open building blocks like Wedge 100 switches and FBOSS, allowing for flexible adaptation to future growth, improved power efficiency, and enhanced reliability through isolated failures.
  44. Global connectivity: Working together to bring more people onlineThis post announces new initiatives and partnerships to expand global internet connectivity. It details field trials for Terragraph technology in Budapest and Kuala Lumpur, in collaboration with Magyar Telekom and Telenor, respectively. It also highlights contributions to the TIP mmWave Networks project group and partnerships with Nokia, Intel, and Radwin for Terragraph-certified solutions. The post also announces the expansion of rural connectivity collaborations with Telefonica in Peru and new initiatives with Orange in Africa, focusing on OpenCellular deployments and new operating models for rural infrastructure. Additionally, it mentions trials of OpenCellular with BT, Nokia, and Cavium in Scotland, and with Vodacom in Africa, and Telenor in Pakistan. The use of machine learning-optimized population density maps for identifying deployment locations and data science methods for optimizing backhaul and access routes are also mentioned.
  45. 2017 Year in review: Better global networksThis post details advancements in Meta's global networking infrastructure in 2017. Key contributions include the deployment of 100G optical connections and transceivers for data center networking, the development of disaggregated networking platforms (Wedge, Wedge 100, Backpack) adopted by the broader community, and the creation of the Express Backbone network for optimized cross-data center traffic. It also highlights the Edge Fabric for steering user traffic and FCR for monitoring network devices. For global connectivity, the post mentions the second test flight of the Aquila solar-powered aircraft, record-breaking wireless data transfer rates (36 Gb/s and 80 Gb/s), and ground-to-air transmission tests. The Terragraph system was tested in San Jose for urban connectivity. Furthermore, Meta open-sourced Open/R, its routing platform, which is now used in backbone and data center networks.
  46. Facebook and Airbus Working Together to Advance High Altitude ConnectivityThis post announces a collaboration between Facebook and Airbus to advance spectrum and aviation policy for High Altitude Platform Stations (HAPS). It highlights the importance of HAPS for bringing broadband connectivity to rural regions and details the ongoing work with international bodies like the ITU and ICAO to secure necessary spectrum and integrate HAPS into global aviation infrastructure. The post emphasizes the need for industry collaboration to demonstrate the viability of HAPS systems.
  47. Open/R: Open routing for modern networksThis post announces the open-sourcing of Open/R, an extensible network routing platform used in Facebook's backbone and data center networks. It highlights Open/R's applicability across different network topologies (WANs, data center fabrics, wireless meshes) and its support for various hardware and software systems. Key features like automatic IP prefix allocation, RTT-based cost metrics, graceful restart, and fast convergence are mentioned. The post details its use in Terragraph, the Express Backbone, and data center fabrics, emphasizing its flexibility and extensibility for rapid innovation in network functions and applications. It also discusses the design principles for modern hardware and software environments, including robust testing and emulation frameworks.
  48. Aquila: What’s next for high-altitude connectivity?This post introduces the Aquila program, focusing on High Altitude Platform Station (HAPS) connectivity as a building block for global internet access. It details the system architecture, including communication payloads, fleets of aircraft, fleet operations, gateways, fiber PoPs, cell base stations, and access points. The post also highlights the regulatory and spectrum challenges that need to be addressed for a commercially viable HAPS connectivity system.
  49. Aquila’s successful second flight: Another step forward in bringing the world closer togetherThis post details the successful second full-scale test flight of the Aquila aircraft, which flew for 1 hour and 46 minutes. Key technical improvements implemented based on lessons from the first flight include the addition of 'spoilers' for landing control, integration of hundreds of sensors for data gathering, modifications to autopilot software and communication radios, application of a smoother finish for improved aerodynamics, and installation of a horizontal propeller stopping mechanism for landing. The flight demonstrated nominal behavior of all systems, a significantly improved climb rate, and provided extensive data on drag, structural response to flight, and spoiler effectiveness. The landing was successful, with the aircraft stopping in approximately 10 meters on a prepared gravel pad, though only one propeller locked horizontally as planned.
  50. Facebook’s new tech at F8 2017This post details advancements in wireless data transfer records (36 Gbps over 13 km with millimeter-wave, 80 Gbps with optical cross-link) and the development of the Tether-tenna, a helicopter-based connectivity solution for disaster relief. It also introduces the x24 and x6 Surround 360 cameras for 6DoF VR experiences and the 360 Capture SDK for capturing and sharing VR content. Enhancements to 360 video encoding and playback include a gravitational view-prediction model and content-dependent streaming. Finally, it announces the open-sourcing of Caffe2, a lightweight deep learning framework.
  51. Connectivity: A building block approachThis post details Meta's 'building block' strategy for global connectivity, focusing on specific technologies for different use cases. For dense urban centers, it introduces Terragraph, a multi-node wireless system extending fiber using wireless, with a pilot deployment in San Jose. It highlights the challenge of line-of-sight obstructions for 60GHz channels and the software developed to route around them in real-time. For rural regions, it discusses MMW and aircraft technology, referencing Aquila (solar-powered UAV) and record-breaking MMW (36 Gbps over 13 km) and optical cross-link (80 Gbps over 13 km) data transfer achievements. It also mentions a ground-to-air MMW record (16 Gbps to a Cessna). For emergency situations, it introduces Tether-tenna, a helicopter-based solution for rapid connectivity restoration.
  52. Designing 100G optical connectionsThis post details the design and deployment of 100G single-mode optical transceivers within Meta's data centers. It highlights the challenges of upgrading to higher data rates, the decision to move to single-mode fiber for future-proofing, and the optimization of transceiver specifications (CWDM4-OCP) to reduce cost and power consumption. The post also discusses strategies for achieving scale on schedule by mitigating technology and deployment risks, and working with suppliers to ensure rapid production ramp-up. This contributes to the broader infrastructure evolution by addressing the critical need for high-bandwidth, efficient data center networking.
  53. This browser tweak saved 60% of requests to FacebookThis post details Meta's collaboration with Chrome and Firefox to significantly improve browser caching efficiency. It explains the technical challenges of revalidation requests and how Meta implemented content-addressed URLs to allow for year-long cache expiration. The post highlights specific browser optimizations, including Chrome's fix for POST requests and reload behavior, and Firefox's adoption of the `cache-control: immutable` header. These efforts resulted in a 60% reduction in static resource requests and improved page load times, contributing to a more efficient web experience.
  54. Legacy support on IPv6-only infraThis post details the technical implementation of supporting IPv4 traffic on an IPv6-only data center infrastructure. It describes modifications to Layer 4 (L4LB/shiv) and Layer 7 (L7LB/proxygen) load balancers, including the use of BGP with ExaBGP, IPv4 link-local addresses as BGP next-hops for IPv6-only servers, IP tunneling of IPv4 traffic within IPv6, the development of the gnlpy library for IPVS forwarding configuration via Netlink, and a custom kernel module for mixed-version IP tunnel decapsulation. It also covers the use of direct server return (DSR) for responses and the configuration of IPv4 default gateways and link-local addresses on servers.
  55. Favorite hacks of 2016This post details several hackathon projects that contributed to Meta's engineering capabilities. The 'Voyager' project developed a prototype transponder for 100G data center infrastructure, capable of transmitting 200GB Ethernet traffic over a single wavelength, and was later contributed to the Telecom Infra Project. Another hackathon project built a private, live-streaming experience within Messenger by leveraging the 'multiway' backend service, exposing it via the Graph API for third-party WebRTC clients, which influenced the 'Parties' feature for Gear VR. The 'Safety Check rich notifications' hackathon project enhanced Safety Check with in-line actions and a Community Help feature for local resources. 'Account Kit instant verification' explored using active Facebook sessions to verify phone numbers for a faster login flow, now available for Android. 'Nuclide Swift' integrated Swift support into the Nuclide IDE. 'Atom in orbit' created a browser-runnable version of the Atom text editor. 'Adaptive attachments' extended the Recommendations feature to enable dynamic content updates in posts for shared albums, crowdsourced videos, and collaborative games.
  56. Reviewing Aquila’s first full-scale test flightThis post details the lessons learned from Aquila's first full-scale test flight, specifically focusing on a structural failure during landing. The failure was attributed to high-than-expected wind conditions lofting the aircraft, an autopilot response that prioritized glidepath over airspeed, and an insufficient drag device. The post outlines two key engineering improvements for future designs: the incorporation of a drag device (spoiler or airbrake) and a revised autopilot command to prioritize airspeed limitation over altitude tracking during descent. It also highlights the ongoing iterative process of design, testing, and learning in developing this experimental aircraft.
  57. Open population datasets and open challengesThis post details the creation and public release of high-resolution population datasets derived from satellite imagery and computer vision techniques. It describes the technical approach of using convolutional neural networks to identify buildings globally, the trade-offs made for scalability, and the collaboration with Columbia University and the World Bank for population estimation and validation. It also presents an analysis of population aggregation near urban areas using clustering algorithms and discusses the implications for designing communication technologies that can bridge distances up to 63 km to connect 99% of the population in analyzed countries. The post highlights the use of these datasets for humanitarian aid and disaster response planning.
  58. Facebook demonstrates record-breaking data rate using millimeter-wave technologyThis post details the demonstration of a record data rate of nearly 20 Gbps over 13 km using millimeter-wave (MMW) technology. It highlights the development of custom-built, high-linearity RF components, high-efficiency power amplifiers using post-amplification spectrum multiplexing, and advanced antenna tracking techniques (sun tracking for bore-sight calibration and conical scanning for precision pointing). The post also discusses the technical challenges of MMW spectrum, including RF component technology, antenna design and tracking, and atmospheric attenuation, and outlines next steps for flight testing air-to-ground bidirectional links.
  59. Introducing Backpack: Our second-generation modular open switchIntroduced Backpack, a second-generation modular open switch platform designed for 100G data centers. Backpack features a disaggregated architecture with switch elements, separation of data, control, and management planes, and an orthogonal direct chassis for improved signal integrity and thermal performance. It supports low-cost 55C optics and integrates with FBOSS and OpenBMC. The internal topology is a two-stage Clos, utilizing BGP for route distribution with fabric cards as route reflectors. The post details engineering validation, design validation, and production validation testing phases, including specific tests for functionality, traffic forwarding, failure recovery, performance, and route scale. It also highlights the integration of BGP monitoring with Scuba and the development of a FIB monitor for data plane visibility. Backpack is being deployed in production pods to meet growing traffic demands.
  60. Introducing Community Cellular Manager: A management and deployment suite for small-scale cellular networksIntroduced the Community Cellular Manager (CCM) open-source project, a software management and deployment suite for operating small-scale cellular networks. CCM enables organizations with limited technical capacity to build these networks using platforms like OpenCellular or third-party RAN solutions. Key features include reliable operation in environments with unreliable power and backhaul, disconnected operation for local voice and SMS, a simple web interface, and local agent management via cloud UI or SMS. The technology consists of a CCM client (for local breakout and disconnected operation) and a CCM cloud (for centralized management). It integrates with third-party radio network stacks (e.g., Osmocom), local voice/SMS services, and persistent storage. Deployments in the Philippines with Globe Telecom and the University of the Philippines validate the model of rural access.
  61. An open approach for switching, routing, and transportIntroduced Open Packet DWDM, a packet transponder and open line transport system combining packet and DWDM technology for fiber optic transport networks. Developed Voyager, an industry-first "white box" transponder and routing solution based on Open Packet DWDM, leveraging data center technologies like Wedge 100's switch ASIC and Acacia Communications' DSP ASIC and optics module. Contributed Open Packet DWDM to the Telecom Infra Project (TIP) via the Backhaul: Open Optical Packet Transport (OOPT) project group to encourage open and programmable network architectures. Open-sourced Voyager software.
  62. Building one of the highest-capacity subsea cables in the PacificThis post details Meta's partnership with Google and Pacific Light Data Communication (PLDC) to build the Pacific Light Cable Network (PLCN), a new subsea cable system. PLCN will be one of the highest-capacity trans-Pacific systems with an initial design capacity of 120 Tbps, connecting Los Angeles and Hong Kong directly. It introduces a flexible approach to subsea cable deployment, separating the wet plant from optical technology, allowing independent equipment upgrades by each party as technology improves. This is the second such project for Meta, following the MAREA cable.
  63. Inside Facebook’s hardware labs: Moving faster with more collaborationThis post details the creation and capabilities of 'Area 404', a new 22,000-square-foot collaborative hardware lab. It highlights the lab's state-of-the-art machine tools (9-axis mill-turn lathe, 5-axis vertical milling machine, water jet, CNC fabric cutter, coordinate measuring machine, electron microscope, CT scanner) and their application in accelerating hardware development cycles from weeks to days. The post showcases how this collaborative space enables faster prototyping and iteration for projects across Infrastructure (e.g., Wedge, 6-pack, Open Vault, Big Sur, TIP components), Connectivity Lab (e.g., Aquila, Terragraph, Project ARIES, optical detectors), and VR (e.g., Surround 360 camera rig, Oculus prototypes).
  64. Flying Aquila: Early lessons from the first full-scale test flight and the path aheadThis post details the first full-scale test flight of Aquila, a solar-powered aircraft designed to provide broadband internet access. It covers early lessons learned from the flight, including successful validation of takeoff procedures, aerodynamic models, battery and power performance, and autopilot functionality. The post also highlights challenges ahead, such as achieving long-duration flight and transmitting wireless signals, and mentions a structural failure experienced just before landing.
  65. Harnessing light for wireless communicationsThis post details the development of a novel luminescent detector for free-space laser communication. The detector uses plastic optical fibers doped with organic dye molecules to absorb blue light over a large area (126 cm²) and re-emit it as green light to a small photodiode. This approach concentrates light without expensive optical components, enabling a larger active area and field of view without sacrificing speed (achieving >2 Gbps). This significantly reduces the need for precise alignment and improves data transmission reliability by collecting more light. The research also explores the potential for higher-speed materials and nanophotonics/plasmonics for future improvements.
  66. Introducing OpenCellular: An open source wireless access platformIntroduced OpenCellular, an open-source, cost-effective, software-defined wireless access platform designed to improve connectivity in remote areas. The platform supports 2G, LTE, and Wi-Fi, and is composed of two main subsystems: General-purpose and Base-band Computing (GBC) with integrated power and housekeeping, and Radio Frequency (RF) with an integrated analog front-end. The hardware design, firmware, and control software are planned to be open-sourced to enable local deployment and operation of wireless infrastructure. The design emphasizes modularity, scalability, low CAPEX/OPEX, simplified installation, and maximization of existing infrastructure.
  67. Growing the Wedge/Wedge 100 communityThis post details the growth of the community around Meta's disaggregated networking hardware (Wedge and Wedge 100) and software (FBOSS). It highlights contributions from AT&T Foundry (TORC, running Docker/Mesos on Wedge), SK Telecom (porting Indigo/OpenFlow and investigating OpenSwitch), Canonical (MAAS/Snappy Ubuntu Core for managing Wedge as a server), and SnapRoute (contributing FlexSwitch, an open-source L2/L3 networking stack). The post emphasizes the increasing adoption and development on Wedge platforms, driven by the open hardware and software building blocks.
  68. Networking @Scale, May 2016 — RecapThis post details presentations from the Networking @Scale 2016 conference, highlighting advancements in large-scale networking. Key contributions include Facebook's deep dives into Terragraph (a 60 GHz wireless system for urban connectivity) and Open/R (a modular routing platform), as well as their experiences scaling Facebook Live with techniques like request coalescing and adaptive bit rate encoding. The post also covers Microsoft's PingMesh and NetBouncer for data center network monitoring, Google's iterations on its B4 SDN WAN, AT&T's use of EVPN for network scaling, Netflix's EVCache for global data caching, and JPL's unique networking challenges for Mars rovers. Day 2 focused on IPv6, with Facebook sharing experiences converting its backbone and data centers to IPv6 and advocating for IPv6-only network support.
  69. Introducing Open/R — a new modular routing platformIntroduced Open/R, a custom-built, extensible distributed network application platform. Open/R was initially designed as a shortest-path routing system for Terragraph, a multi-node wireless network for urban internet connectivity. Its modular design allows for additional applications beyond basic routing, and it has been adapted for other networking infrastructure. Open/R enables faster prototyping and deployment of new network applications compared to industry standards, contributing to improved speed, efficiency, and quality of internet connectivity. The platform leverages Thrift for message encoding and ZeroMQ for message exchange, with components like KV-STORE, LINK MONITOR, DECISION (using Dijkstra SPF and computing loop-free alternatives), and FIB (interacting with routing hardware via FibAdapter and FibAgent). It also incorporates fast failure recovery mechanisms and has been tested to scale to networks with thousands of nodes. The company plans to contribute Open/R to the open source community and the Telecom Infra Project (TIP).
  70. Introducing Facebook’s new terrestrial connectivity systems — Terragraph and Project ARIESThis post introduces two new terrestrial connectivity systems: Terragraph, a 60 GHz multi-node wireless system for dense urban areas leveraging WiGig standards and optimized for low-cost production, and Project ARIES, a proof-of-concept Massive MIMO test platform with 96 antennas demonstrating high spectral and energy efficiency for rural connectivity. Terragraph utilizes IPv6-only nodes, an SDN-like cloud compute controller, and a re-architected TDMA-TDD MAC layer for improved efficiency. Project ARIES focuses on spatial multiplexing to serve multiple streams over the same spectrum, aiming for over 100 bps/Hz spectral efficiency. Both systems are being tested and prepared for broader trials and open contribution to the ecosystem.
  71. How we built Facebook Lite for every Android phone and networkThis post details the architecture and engineering decisions behind Facebook Lite, a version of Facebook optimized for emerging markets with low-bandwidth networks and low-spec Android devices. Key technical contributions include a proxy server architecture with a thin client VM, a custom message protocol over TLS to minimize data usage and improve performance on 2G networks, dynamic shared dictionary compression, LZMA2 and DEFLATE compression, image optimization by serving exact-size images from a dedicated server, and a diff mechanism for sending incremental updates. The design also prioritizes minimizing CPU, storage, and RAM usage on the client to ensure compatibility with older devices like Gingerbread-era phones.
  72. Connectivity Lab custom-designs two-axis gimbal for air-to-air and air-to-ground laser communicationsThis post details the custom design and development of a lightweight, two-axis gimbal for UAVs to support air-to-air and air-to-ground laser communications. It describes the design considerations, including swiveling, weight, and aerodynamic appropriateness, and presents various design configurations considered, culminating in the final carbon fiber and magnesium composite gimbal weighing approximately 3.5 kilograms, which is at least two times lighter than existing gimbal technology.
  73. Adopting an open approach to global networks with the Telecom Infra ProjectThis post announces the formation of the Telecom Infra Project (TIP), an engineering-focused initiative to drive the development of next-generation open network infrastructure components. It highlights the need for collaboration to address the challenges of increasing global data traffic and the limitations of monolithic, costly network evolution. TIP aims to produce simple, efficient, and flexible technologies by working in the open, similar to the Open Compute Project (OCP) but focusing on interoperable software systems and components for access, backhaul, and core networks. The post also touches upon advancements in data center optical networks and the benefits of IPv6 for future network design.
  74. Connecting the world with better mapsThis post details the development of a high-resolution (5-meter) population distribution dataset for 20 countries by applying computer vision techniques to satellite imagery. The process involved identifying human-built structures as proxies for population density, using a combination of image preselection, a modified Facebook image-recognition engine (deep convolutional neural network), and a weakly supervised neural network. The resulting data was used to redistribute census counts, providing granular insights for planning short-range access networks and backhaul solutions for internet connectivity in underserved areas. The scale of the operation involved processing 14.6 billion images on thousands of servers.
  75. Open networking advances with Wedge and FBOSSThis post details the introduction and scaling of Wedge, a disaggregated top-of-rack network switch, and FBOSS, its Linux-based control software. It covers the engineering effort behind Wedge's open hardware design, the operationalization of Wedge and FBOSS through custom deployment tools (fbossdeploy) and monitoring strategies, the lessons learned from treating switches like servers (managing service infrastructure rollouts and avoiding circular dependencies), and CPU protection mechanisms to prevent control plane overload. It also looks ahead to Wedge 100 and the 6-pack aggregation switch.
  76. Milestones and plans for the OCP Networking ProjectThis post details the progress and future plans of the OCP Networking Project, co-led by Facebook. It highlights the evolution of open hardware and software in networking, from initial concepts to adopted specifications and products like FBOSS and OpenBMC. The project's goal is to increase innovation by encouraging network disaggregation and the development of open source hardware and software, contributing to the broader goal of improving internet infrastructure and access.
  77. Building for emerging markets: The story behind 2G TuesdaysThis post introduces '2G Tuesdays,' an internal initiative to simulate 2G network conditions for employees. This aims to foster empathy and understanding of user experiences in emerging markets, complementing existing tools like Connection Class and Augmented Traffic Control by focusing on product decision-making and user pain points on low-bandwidth networks.
  78. Improving the Linux kernel with upstream contributionsThis post details Meta's "upstream-first" philosophy for contributing to the Linux kernel, focusing on improving storage and networking stacks. Key contributions include enhancements to blk-mq for block storage performance, stabilization of Btrfs for production data protection, implementation of the New Vegas congestion control algorithm, writeback support for cgroups for fine-grained IO limiting, advancements in network virtualization (GUE, IPv6 flow label, ILA), IPv6 performance improvements by scaling the routing tree, and a caching layer for RAID 5/6 to address the write hole and improve performance. The post also highlights progress in release engineering for faster kernel deployments.
  79. IPv6: It’s time to get on boardThis post details Meta's migration to IPv6, highlighting its technical benefits such as easier network management, end-to-end connectivity integrity, and improved interoperability. It reports a 10-15% performance improvement for accessing Facebook over IPv6 and advocates for broader adoption due to the exhaustion of IPv4 addresses.
  80. Building communications networks in the stratosphereThis post details the development of Aquila, a high-altitude, long-endurance solar-powered aircraft designed to provide a 50-km communications radius for up to 90 days. It describes the aircraft's architecture, including its large, lightweight design (Boeing 737 wingspan, one-third the weight of an electric car), monocoque carbon fiber construction, and flight profile (60,000-90,000 feet) optimized for solar charging and energy conservation. It also announces advancements in free space laser communications, achieving data rates of tens of Gbps, leveraging technologies from Meta's data centers.
  81. F8: Big technology bets and open source announcementsThis post announces the first test flight of Meta's large UAV platform, a significant milestone in their efforts to provide internet connectivity to remote communities. It also highlights ongoing work with satellites, cost-effective base stations, and free-space optics as part of their broader connectivity initiatives.
  82. Improving Facebook on AndroidThis post details significant efforts to improve the Facebook Android app's performance, data efficiency, networking, and application size, specifically targeting users in emerging markets with limited bandwidth and device capabilities. Key contributions include optimizing app startup times, reducing image data usage through WebP and adaptive resolution loading, adopting OkHttp for improved networking, and reducing app size by leveraging Google Play's multi-APK feature. These improvements directly support the goal of connecting the next five billion users by making Facebook more accessible and usable on a wider range of devices and network conditions.
  83. Introducing “Wedge” and “FBOSS,” the next steps toward a disaggregated networkThis post introduces 'Wedge,' a disaggregated top-of-rack network switch, and 'FBOSS,' a Linux-based operating system for it. These innovations aim to make network infrastructure more flexible, scalable, and efficient by separating hardware from software, mirroring the disaggregation principles applied to servers and storage through the Open Compute Project. This represents a significant step in building more modular and controllable network components, aligning with the broader goal of expanding internet access through robust and adaptable infrastructure.
  84. Announcing The Connectivity Lab at FacebookThis post announces the formation of the Connectivity Lab at Facebook, a team dedicated to developing new aerospace and communication technologies to advance the Internet.org mission of improving and extending internet access. The lab will focus on new delivery platforms, including planes and satellites, and brings together experts from Ascenta, NASA's Jet Propulsion Laboratory, NASA's Ames Research Center, and the National Optical Astronomy Observatory.
  85. Under the Hood: Building the Location APIThis post details the evolution of Facebook's location services, focusing on the technical challenges and solutions for making location tagging universal, building a global places directory, and integrating location data into the Timeline map. Key technical contributions include: a new indexing system for places using four polygons and name prefixes to handle a global dataset, a ranking system for places based on features like check-ins and opening hours, infrastructure to farm out data fetching for the Timeline map to multiple servers, and a deterministic aggregation algorithm for map pins based on closest neighbor relationships. It also describes the development of systems to suggest locations for existing albums and the launch of APIs for third-party applications to post content with location tags and access place search.