Globalping: A Community-Driven, Open-Source Platform for Scalable, Real-Time Network Measurements

This paper introduces Globalping, an open-source, community-driven platform that democratizes access to scalable, real-time global network diagnostics by leveraging a distributed network of user-hosted probes and offering seamless integration with tools like Slack and GitHub.

The Gist

Imagine the internet as a massive, global city with billions of roads, intersections, and buildings. Sometimes, traffic jams happen, or a bridge collapses, and you need to know exactly where the problem is and how bad it is.

For a long time, checking these "internet roads" was like trying to get a private jet to fly over the city just to look at a traffic jam. Only a few experts with expensive equipment and special clearance could do it. This is what tools like RIPE Atlas were like: powerful, but hard to access and often slow to give you an answer.

Globalping is the paper's new idea: a "crowdsourced traffic drone network" that anyone can use.

Here is the breakdown of how it works and what the paper found, using simple analogies:

1. The Core Idea: A Global Network of Volunteers

Instead of building expensive, permanent hardware stations in every country, Globalping asks regular people and companies to install a tiny piece of software (a "probe") on their own computers or servers.

• The Analogy: Think of it like a neighborhood watch. Instead of hiring a private security firm, you ask neighbors to keep an eye on the street. Globalping turns thousands of these "neighbors" into a global team that can instantly check the internet from their specific location.
• The Tech: These probes run in "containers" (like lightweight, portable boxes that can be dropped onto any computer). This makes them super easy to set up, unlike the heavy, custom hardware required by older systems.

2. What It Actually Does

The platform lets you ask three main questions about the internet, and it answers them in real-time (instantly, as the data happens):

• Ping: "Is the destination open, and how long does it take to get there?" (Like shouting across a canyon to see if someone hears you and how long it takes for the echo to return).
• Traceroute: "What path did the message take?" (Like tracing a package's journey through every warehouse and truck it passed to see where it got stuck).
• DNS Lookups: "Can we find the address of this website?" (Like checking the phonebook to see if a number is valid).

3. The "Superpowers" (New Features)

The paper highlights that Globalping does things older tools don't do well:

• Real-Time Streaming: Older tools give you a snapshot every few minutes (like a daily weather report). Globalping gives you a live video feed of the network status.
• Integration: It plugs directly into tools teams already use, like Slack (for team chat) and GitHub (for software code).
  • Analogy: If a server goes down, Globalping doesn't just send you an email; it posts a message directly in your team's chat room saying, "Hey, the road to the server is blocked," so you can fix it immediately.
• Democratization: You don't need to be a network engineer to use it. The interface is designed so a beginner can run a test as easily as sending a text message.

4. The "Test Drive" (User Study Results)

The authors didn't just build it; they tested it against the old standard (RIPE Atlas) with 40 regular people (not just experts).

• The Task: They asked everyone to run a simple "Ping" test (checking if a website is reachable).
• The Speed:
  • Globalping users finished the task in about 3 minutes.
  • RIPE Atlas users took nearly 7 minutes (more than double the time).
• The "Ease of Use" Score (SUS):
  • They gave the tools a report card on how easy they were to use.
  • Globalping: 65.62 (Above average, considered "good").
  • RIPE Atlas: 42.38 (Below average, considered "poor").
  • The Takeaway: Regular people found Globalping much less confusing and frustrating to use.

5. Safety and Rules

The paper explains that they built strict rules to prevent abuse.

• No Botnets: These probes can only do specific, safe checks (like a ping). They cannot be used to launch cyberattacks or flood websites with traffic.
• One-Way Street: The probes only talk to the Globalping server; they never accept incoming connections from strangers, making them very secure.

Summary

The paper presents Globalping as a modern, community-powered tool that makes checking the health of the internet as easy as checking the weather. It is faster, easier to use, and integrates better into daily work than older systems, allowing both tech experts and regular users to spot and fix internet problems instantly.

Technical Summary

1. Problem Statement

The paper identifies a critical gap in the current landscape of Network Diagnostics (ND) and monitoring tools:

• High Barrier to Entry: Existing tools (e.g., RIPE Atlas) are often designed for technical professionals, requiring complex configurations, specialized hardware, or deep networking knowledge, making them inaccessible to non-technical users and small businesses.
• Latency in Data Delivery: Most platforms provide "near-real-time" data, which involves discrete delays due to scheduling, batching, or polling. This is insufficient for scenarios requiring immediate awareness or rapid response to transient network issues (sub-second latency).
• Lack of Integration: Major ND platforms are not natively embedded into modern software development lifecycles (CI/CD pipelines like GitHub Actions) or communication workflows (Slack, Discord), forcing network measurement to remain a standalone research task rather than an integrated operational tool.
• Hardware Dependency: Traditional solutions often rely on stable, long-term physical hardware probes, which are difficult to deploy and scale compared to ephemeral, containerized infrastructure.

2. Methodology

The authors employed a multi-faceted approach to design, implement, and evaluate Globalping:

• System Design & Architecture:

  • Architecture: A monolithic Core API Server coupled with a Probe Management Module, utilizing a globally distributed network of user-hosted probes.
  • Infrastructure: Probes are containerized using Docker (or Podman), allowing them to run on any server with minimal resources (1 vCPU, 512 MB RAM).
  • Communication: Uses WebSockets for real-time, bidirectional communication between the Core API and probes, enabling continuous data streaming with sub-second latency.
  • Data Management: Redis is used for caching, session management, and storing measurement results with Time-To-Live (TTL) settings to handle concurrency.
  • Security: Probes are restricted to predefined commands (ICMP, DNS, limited HTTP) and cannot act as general traffic generators. Rate limiting, domain blacklisting, and strict parameter bounds prevent DDoS abuse and unauthorized scanning.

• User Study (System Usability Scale - SUS):

  • Participants: 40 participants recruited via Prolific with random backgrounds (no technical filtering) to represent a broad user base.
  • Task: A single, fundamental task: executing a ping measurement from a location in Germany to example.com.
  • Comparison: Participants were split into two groups (Globalping vs. RIPE Atlas) to compare usability, time-to-completion, and success rates.
  • Metrics: SUS scores (0-100 scale), task completion time, and binary success/failure rates.

• Comparative Analysis:

  • The authors compared Globalping against industry tools (RIPE Atlas, PerfOps, Cedexis, KeyCDN Tools) based on features like official integrations, containerization support, free tier policies, and target audience accessibility.

3. Key Contributions

• Democratization of Network Diagnostics: Globalping lowers the barrier to entry, allowing non-technical users, companies, and experts to perform diagnostics (ping, traceroute, DNS, HTTP) via an intuitive UI, CLI, or REST API without creating an account for basic usage.
• Real-Time Streaming Architecture: Unlike the "near-real-time" polling of competitors, Globalping utilizes WebSockets to deliver measurement results continuously with sub-second latency.
• Official Containerization: Globalping is the first major platform to officially support and document the deployment of probes via Docker containers, significantly simplifying setup compared to hardware-based alternatives.
• Native Workflow Integrations: It offers official, maintained integrations with Slack and GitHub (including GitHub Actions), allowing network diagnostics to be embedded directly into collaborative workflows and CI/CD pipelines.
• Economic Model: A credit-based system where users earn credits by hosting probes (or donating) and spend them to run tests, fostering a self-sustaining community ecosystem.

4. Results

• Usability (SUS Score):
  • Globalping: Achieved a SUS score of 65.62 (approaching the 50th percentile of 68).
  • RIPE Atlas: Achieved a SUS score of 42.38 (well below the 10th percentile of 51).
  • Statistical Significance: A paired t-test yielded a p-value of 0.00057, with a Cohen's d of 1.64 (a large effect size), indicating Globalping is significantly more usable.
• Task Efficiency:
  • Time: Globalping users completed the ping task in an average of 3.23 minutes, compared to 6.87 minutes for RIPE Atlas users.
  • Success Rate: Only 3 out of 20 Globalping users failed to complete the task, whereas 5 out of 20 RIPE Atlas users failed.
• Adoption & Scale:
  • As of the study, Globalping had 3,032 probes across 632 cities in 114 countries.
  • The system processes over 300,000 measurements daily with an average API response time of less than 14 milliseconds.

5. Significance

• Paradigm Shift in ND: Globalping moves network diagnostics from a specialized, research-centric activity to an accessible, real-time utility for the general public and DevOps teams.
• Operational Integration: By integrating directly into Slack and GitHub, it enables "shift-left" network monitoring, allowing teams to detect and troubleshoot connectivity issues immediately within their existing collaboration tools.
• Scalability via Ephemeral Infrastructure: The containerized approach allows for rapid scaling and geographic diversity without the logistical burden of deploying physical hardware, addressing the bias often found in hardware-based probe networks.
• Future of Network Monitoring: The paper demonstrates that open-source, community-driven models can outperform established, research-focused platforms in terms of usability and accessibility, suggesting a future where network health monitoring is a standard, integrated feature of software development and business operations.

Limitations Noted: The study relied on non-technical participants, which may have skewed the SUS scores lower than they would be for engineers. Additionally, Globalping currently has fewer probes (3k) compared to RIPE Atlas (12k), though the lower barrier to entry is expected to accelerate its growth.