Show HN: NetFabric – next-gen network monitoring solution

netfabric.ai

6 points by pentab a month ago

Hi HN! We’re Tobias and Beni, co-founders of NetFabric. NetFabric is a next-gen network monitoring solution that’s here to tackle the chaos of network downtime with a fresh approach using mathematical models and machine learning: We integrate all your network data sources into a single, easy-to-use platform, giving you clear answers to any network question in real time.

# Why Network Monitoring Needs a Revolution Network downtime is a massive headache for companies everywhere. Just an hour of downtime can cost millions—Amazon lost $34 million in just one hour in 2021! But the pain doesn’t stop at the financial hit. Downtime disrupts operations, ties up engineers in endless troubleshooting, and damages customer trust. The issue with current network monitoring tools is that they offer only basic insights into the complex world of networks. They rely on fragmented data, leading to false positives and incomplete problem resolution. This often means engineers spend excessive time deciphering alerts and dealing with misleading signals, which only prolongs downtime and worsens operational inefficiencies.

# Our Innovative Approach Leveraging Years of Research At NetFabric, we aim to revolutionize the network monitoring world with a bold new approach. Unlike traditional solutions, NetFabric seamlessly merges observations coming from a wide range of network sources (such as routing information, forwarding tables, device configurations, and logs) into one coherent knowledge base. As a result, it delivers actionable intelligence, offering a comprehensive and accurate understanding of network issues while effectively managing the flood of alerts from conventional tools.

NetFabric achieves this by addressing two fundamental challenges: 1. Network Diversity: First, real-world networks vary widely in terms of deployed devices, protocols, and architecture; this network diversity causes variations in the representations of monitoring data (such as configurations, metrics, or logs), preventing one-size-fits-all monitoring solutions. NetFabric addresses this by integrating large language models (LLMs) to handle non-standard data sources and user queries effectively. 2. Protocol Complexity: Second, at their core, networks are driven by intricate protocols. With a rigorous understanding of protocol logic, network monitoring can generate valuable insights from the available data, e.g., by explaining why a given network problem was observed, and how the problem can be resolved. Since such a systematic understanding of protocols is beyond LLMs, NetFabric complements LLMs with advanced mathematical models informed by a decade of academic research. By merging these technologies, NetFabric delivers a unified and complete view of complex networks and provides actionable insights while cutting down on false positives. This dual approach not only optimizes resource allocation but also helps organizations minimize downtime and keep their operations running smoothly.

# Some Of The Use Cases That We Enable: - Faster issue resolution: We naturally guide operators towards the problem across data sources and layers, enabling real-time root cause analysis. - Actionable alerts & reports: We make alerts actionable by adding the missing context to correlate, aggregate and filter them before they reach your inbox. - What-if & resilience analysis: We improve availability and proactively prevent incidents by leveraging our models to evaluate what-if scenarios. - Meaningful AIOps: We unlock large language models (LLMs) for network data, and combine them with our mathematical models where their reasoning abilities fall short. - Many other use cases: Furthermore, our approach applies to network security, application & multi-cloud monitoring, planning, cost management, self-healing networks, auto-documentation, and more.

We’d love to hear your thoughts on network monitoring, share your experiences, and get your feedback!

KomoD a month ago

Please read the Show HN rules.

https://news.ycombinator.com/showhn.html

> Off topic: blog posts, sign-up pages, newsletters, lists, and other reading material. Those can't be tried out, so can't be Show HNs. Make a regular submission instead.

I see no way to try this, I only see a contact form.

(I can't even figure out what it's supposed to be, there's so many buzzwords)

  • pentab a month ago

    Thanks for your recommendation and sorry for not fully understanding the rules. I tried to fix this but did not see how to edit my submission.

abpavel a month ago

Hi Tobias and Beni, Pavel here, co-founder of IP Fabric, the Automated Network Assurance Platform for the Enterprise Networks. We're solving for the problem that you're describing, albeit in a more deterministic way, by modeling the entire enterprise network infrastructure. There are also competitors in the valley already - Forward Networks. While we welcome the competition, and while I am curious how you are solving for such a noise coming from the variety of networking products, I must say that you could have chosen a name that is copying us a little bit less.

  • pentab a month ago

    Hi Pavel. Thanks for your comment and for sharing your perspective! We indeed believe both IP Fabric and ForwardNetworks follow promising approaches. A key difference is that we are focusing on a broader range of data sources (including NetFlow, SNMP, BGP, etc) and therefore can also support a live view of a given network.

    Regarding the variety, we agree that network data sources suffer from a lot of subtle variety, which can be really frustrating especially because this variety makes it hard to access very valuable information. This is our main motivation to integrate LLMs, as a powerful tool that can naturally handle subtle variety.

    Regarding the name, it was not our intention to copy or imitate. We chose "NetFabric" because we felt it accurately represented our vision of creating a seamless, integrated network monitoring solution that weaves together diverse network data sources.