Rick Mur

Limitless Networking

Transform ISP Choice from Anecdote to Evidence

Have you ever had to defend your choice of internet service provider? All you can say is: “Everyone says they’re reliable”, “My buddy recommended them.”, “They are the incumbent player”. But when pressed about frequent connectivity issues, then what? Sound familiar? This plays out in businesses across the world every day. We make one of our most critical infrastructure decisions based on hearsay, marketing promises, and gut feelings.

At GNX, we’ve learned that treating ISP selection like choosing a restaurant based on Google reviews is a recipe for disaster. Your internet connection isn’t just a utility anymore. It’s the lifeline that determines whether your video calls freeze during client presentations, whether your cloud applications respond instantly, and ultimately whether your business operates at peak efficiency or struggles with digital friction.

The Hidden Cost of Anecdotal Decisions

When businesses rely on word of mouth or flashy marketing materials, they often discover painful truths too late. The FCC’s Measuring Broadband America program has consistently found that for many ISPs, the measured speeds experienced by subscribers either nearly met or exceeded advertised service tier speeds, but internet usage has fundamentally changed since 2011. What worked for your business three years ago might not serve your current needs.

The real problem isn’t just getting less speed than advertised. It’s discovering that your ISP performs poorly during the hours that matter most to your business, or that their latency to critical cloud services makes your applications feel sluggish despite seemingly adequate bandwidth.

Building Your Evidence Base

Smart ISP selection starts with data collection. Here’s some examples how to gather the evidence you need:

Performance Monitoring Tools Modern speed testing goes beyond simple download speeds, incorporating upload speed, latency, and jitter measurements. Tools like Netflix’s speed test (Fast.com), TestMy.net’s latency testing, and specialized business monitoring platforms provide comprehensive baseline measurements. But don’t stop at a single test. Network stability monitoring tools can run continuously to detect latency spikes and packet drops that occur during peak usage periods.

Real World Performance Data The ACCC’s broadband monitoring program measures real speeds from more than 1000 homes across Australia, providing actual performance data rather than theoretical maximums. Similar programs exist globally. In the US, the FCC’s broadband performance reports offer invaluable insights into how different providers perform under real world conditions. Commercial data sources are also available. Also don’t forget that LLM like Claude and ChatGPT also have a breath of knowledge about this topic and can perform some research for you!

Beyond Maximum Bandwidth Here’s where most businesses get it wrong. They focus solely on advertised speeds without considering performance patterns. Latency under load has become increasingly important as consumers use more roundtrip latency-sensitive real-time applications such as videoconferencing. A connection that delivers 100 Mbps at 3am but drops to 20 Mbps during business hours isn’t truly a 100 Mbps connection for your purposes.

Critical Metrics That Matter

Latency to Your Destinations Don’t just test generic latency. Measure response times to the specific cloud services, data centers, and applications your business depends on. TestMy Latency uses TCP rather than ICMP, providing a more accurate picture of real-world application performance. A provider might have excellent latency to general internet destinations but poor connectivity to your critical AWS region.

Time-Based Performance Patterns ISP performance monitoring should extend to various network locations and simulate traffic exchanges at rapid intervals to ensure thorough assessment. Document how performance varies throughout your business day. An ISP that delivers consistent performance during your peak hours is worth more than one offering higher theoretical speeds with significant variability.

Packet Loss and Jitter These problems can significantly impact gaming and voice chat applications, which usually don’t resend lost information. For businesses using VoIP, video conferencing, or real-time collaboration tools, packet loss and jitter matter more than raw bandwidth.

Making Evidence-Based Choices

Armed with this data, you can make informed decisions that align with your actual needs rather than marketing promises. Consider a recent customer who was choosing between two fiber providers. Provider A offered 500 Mbps for $200 monthly, while Provider B offered 200 Mbps for $180. Based on speed alone, Provider A seemed better.

However, our analysis revealed Provider B had 40% lower latency to their primary cloud applications, maintained 95% of advertised speeds during business hours (versus 70% for Provider A), and showed zero packet loss in evening testing. For their use case involving real-time collaboration tools and cloud-based ERP systems, Provider B delivered superior actual performance despite lower theoretical speeds.

The Path Forward

Transforming ISP selection from anecdote to evidence requires time and effort upfront, but the payoff is substantial. Businesses increasingly depend on robust online operations, making ISP selection a critical decision that can make or break a company.

Start by establishing baseline measurements of your current connection, identify your critical applications and their performance requirements, and gather comparative data on available providers. Test during your actual business hours, not just when it’s convenient. Document everything.

The goal isn’t finding the fastest ISP or the cheapest one. It’s finding the provider whose actual, measured performance best supports your business operations. In a world where digital infrastructure determines competitive advantage, that evidence-based approach makes all the difference.

Your internet connection is too important for guesswork. Make the choice that data supports, not the one that sounds good in a sales presentation.

How GNX+ Transforms the Decision Process

At GNX, we’ve built our platform around this evidence-based philosophy. Our GNX+ algorithm continuously ingests performance data from hundreds and hundreds of ISPs across multiple markets, analyzing real-world metrics including latency patterns, throughput consistency, packet loss rates, and service reliability during peak business hours.

Rather than relying on marketing claims or theoretical speeds, our system processes data points from actual customer experiences, monitoring programs, and independent testing sources. The algorithm weighs these factors against your specific business requirements to recommend ISPs that deliver measurable performance for your use case.

This data-driven approach has helped our customers avoid costly mistakes and identify providers that truly excel where it matters most. Because when you’re making infrastructure decisions that impact your entire operation, evidence beats anecdotes every time.

Your internet connection is the foundation of your digital operations. Choose wisely, choose with data.

Blog Reboot

When I first launched this site, many years ago, it served as a humble lab notebook and sharing short personal stories from my working life. I shared diagrams, Junos configs , and field notes written after late night maintenance windows or proof of concepts.

Those stories took on a life of their own. They brought speaking invites, consulting work, and, most important, a worldwide group of curious friends. Then GNX happened. In a few coffee fueled years we moved from a whiteboard idea to a platform that sources connectivity in nearly every country in the world and a serious company with an amazing group of people. In between fundraising, product build cycles, and the arrival of my son Senna, long form writing became endangered. The blog slipped into hibernation. Four years have passed since the last proper post, which feels like a century in internet years.

Why press publish again

  • The industry needs clarification. We swim in recycled press releases and messaging that skip the trade-offs.
  • My inbox overflows with the similar questions from customers and partners. Turning answers into articles scales better than copy and paste.
  • Writing forces clarity. Every decision at GNX improves after I explain it to a blank page.
  • I simply miss it. Crafting an argument and seeing readers respond is unmatched.

What stays on the GNX domain

The GNX site remains the home for connectivity related blogs, updates, and detailed white papers aimed at procurement teams, network architects, and IT managers.

What lives here

Everything that needs a personal voice: CTO commentary, predictions, behind the scenes lessons, and the occasional rant.

Some ideas where the early drafts are being written:

  • AI with accountability: keeping humans in the loop without slowing progress
  • Transform ISP choice from anecdote to evidence
  • Scoring vendors on real performance data and why that is harder than it looks
  • The hidden cost of cloud front ends and what connectivity providers can do
  • Lessons from early Starlink rollouts
  • Running a distributed engineering team while preserving culture

Coming up next

Expect one blog every few weeks, matching our GNX build cycle rhythm. When a post is live I will craft a condensed LinkedIn note. Social platforms serve as rails, the blog remains the default gateway. Algorithms change, RSS endures. By keeping the canonical version here I ensure links never break.

Say hi!

If you meet me at Cisco Live, another conference, or a toddler swim class, say hello. Bring your toughest networking puzzle or your best trick for a toddler at a festival and we will trade notes.

Thank you for returning after the silence. The modem lights are blinking once more. Let us build a faster, more transparent internet together, one story at a time.

JNCIE-DC Lab Experience

After plenty of hours of studying and labbing the wide ranging topics on the JNCIE-DC blueprint, I took the JNCIE-DC lab exam and passed! I can proudly say I’m JNCIE-DC #389. In this conclusion of the previous JNCIE-DC blogs about my lab setup and about the remote lab environment, I will talk about my experience of taking the lab and how I’ve prepared.

Resources

As I’m working with different cloud provider customers in my daily job it really helped in preparing for this test. Being exposed to the platforms and technologies used gave me a head start, but after really starting to lab the different topics I really felt I had a long way to go. I was also a champion in delaying my actual start of preparation. After committing to the goal of passing the test in Q1 of 2021 I really picked up the pace and started studying multiple nights a week and came into the flow I like to be when preparing for a lab exam.

JNCIE-DC self study workbook

The Juniper Education self-study workbook for the JNCIE-DC lab is by far my most used resource for the lab. This will prepare you for 95% of the test. Note that it does not cover 100% of the topics and you will need to read documentation and have a deep understanding of the technologies on the blueprint of the test.

Examples of this are working with Class of Service configuration on QFX5100 platforms, as the lab only works on vQFX devices which are based on a virtual Q5 PFE rather than a Trident2 chip as used in the QFX5100. This means you really have to be aware of the physical limitations of the platform when applying Class of Service configuration.

The same goes for Virtual Chassis Fabric. Again a feature only supported on QFX5100 (and QFX5110/EX4300), not on the vQFX.

When working through the labs of the workbook you will find that some solutions will not have much explanation other than the correct configuration. Sometimes this made me doubt my own solution of which I was sure was correct as well. Then it really helps by taking the documentation and checking it by yourself to know if your variant is just as right as the configuration in the workbook is.

EVE-NG

My Dell R730 server running bare-metal EVE-NG Pro was a life saver during the preparation. The performance was amazing and I could spin up as much virtual resources as I wanted to test various topologies.

Juniper Documentation

I did not read books as a preparation, but you do need a solid understanding of routing protocols, MPLS and EVPN. To brush up on all the knobs and features I read parts of the Juniper documentation which has excellent sample configurations explaining what’s needed and why.

Taking the lab

As I took the lab in March of 2021. The Juniper offices are closed because of COVID-19 and the lab exam is conducted remotely. I’m fortunate to have a dedicated home office room that I could close and be isolated, as is required according to the lab policy. For all other requirements that your PC and room have to comply with can be found in my previous blog about that.

First be prepared that the JNCIE-DC lab in 2021 is still an 8 hour exam and not a 6 hour exam that other tracks already are.

Early morning on the day of your lab, you will receive an e-mail with an attachment for the Secure Exam Browser. This is a kiosk application that ensures a fully isolated environment where you will access the lab task document, documentation and a remote desktop environment with access to the equipment. Note that you are not able to print the tasks or diagrams of the lab so if you have a widescreen or a dual monitor setup that is a big plus! On my ultrawide monitor I could fit the topology diagram, tasks and remote desktop with notepad and securecrt open next to each other.

You will join a Zoom call where you will need to have (and keep) your webcam enabled as well as your microphone (do not go on mute). On the time mentioned in the e-mail your proctor will join the Zoom session and put you in a separate room where only you and the proctor are in. You will see the proctor jump in and out of your room at random times to check on what you are doing via your webcam, audio and the screenshare that is constantly running.

To access the devices you will open SecureCRT and see sessions to all devices both over out of band management (SSH) and over Console. The sessions will even automatically login so Juniper makes your life easy there. I did have to get used to some latency on the sessions as it felt like the remote desktop was hosted in the US and I’m based in Europe, so having a 150ms-ish latency on the connection does need some adjusting when you are typing fast.

If you want a break to grab a drink or go to the toilet, raise your hand on the Zoom call to attract the attention of the proctor and ask for permission before leaving your desk. The official time for lunch is 1 hour, but you are not allowed to be away for that time of course. The proctor will ask you to eat your lunch in front of the camera at your desk and you can take some time to rest. My proctor was kind enough to allow me to get back to work after about 20 minutes and gave me my official time to finish the exam.

Time management

Managing your time is an absolutely crucial part of taking a lab exam. During my preparation I really struggled initially to finish a decent sized lab on time, but after pushing hard on speed and concentrating I was able to finish the super labs in the workbook in under 4 hours. When I take the lab I’m always very aware of the time I spend on each section or part of the test. I keep track of that by putting timestamps in a text file and which part I finished at that time. The way I like to work is go trough the tasks as quick as I can and finish them so I have a lot of time left for reviews.

When I returned to work after lunch I was about 80% done with the tasks and could finish all of them in about 30 minutes so I had plenty of time left. Now I have to admit when I do these lab exams, my typing speed has increased a lot over my normal speed as I’m trained to work on these configurations for hours. With over 3 hours left on the clock I went back to the very first task and I try to read it like I have never read it before to try and trick myself into coming up with the solution again. Overall I made 2 small changes to the configuration that I don’t know if they made an impact, but to me they felt like the better answer to the task.

With about 30 minutes left I felt like I checked everything twice and did a calculation on how many points I was certain of and how many I had some kind of doubt about. The passing score is not known, but I assumed it’s around 78-80%. With having more then enough points based on my own verification I told the proctor I was finished.

Conclusion

The official waiting period is ‘up to 5 working days’. For my result to come in it actually took those 5 working days so a full week after my lab attempt I received the e-mail that I passed!

This JNCIE-DC exam is definitely a tough one and I feel lucky I was able to pass this one at first attempt. The complexity lies mainly in the fact that everything builds on top of each other and a mistake in one of the layers impacts reachability across pretty much the entire exam. It’s absolutely crucial to be very exact on details of the tasks, but also to not overthink them! Do exactly as the task says within the constraints you are free to choose the solution you want. The exam is not there to trick you, but just to test you on you being an expert!

« Older posts

© 2025 Rick Mur

Theme by Anders NorenUp ↑