Network Outages

Picture of Adam Bahret
Adam Bahret

In our connected world, we have grown accustomed to networks that we rarely give thought to their reliability until there is an outage. In recent days, there has been news of network outages from NVIDIA GeForce now, Meta, and PlayStation. I want to shed some light on the reliability targets that these networks have.

A simple cloud network is designed with a Reliability target of 99.99% (four 9s). That’s on the lower end of the spectrum. Vital networks target even higher – five 9s or 6 9s. Even with these targets, outages occur. It goes to tell us the role probability plays a role in real-world usage. We cannot design a product that lasts reliably all the time. A balance of risk acceptance and design is required. The struggle to achieve this balance is the heartbeat of Reliability Engineering.

Compare the network target numbers to your products. Network designers use reliable components, ample redundancy in hardware and controls, real-time monitoring, and built-in diagnostics to improve performance. What among these methods are you using in your product design?

They also constantly test for interaction failures – issues that come up when an additional system is connected to the network. A Subsystem on its own will be reliable but will cause failures when connected to another subsystem. Are you examining how interactions play a role in introducing failure modes for your product? Recently, real-time monitoring has enabled to predict the outages. Is Prognostics a viable option to consider?

I hope these open-ended questions have planted sufficient seeds for you to examine your product. Let me know if you have any questions about the above ideas or their applicability to your product. Happy to support you.

Share this post