Cable faults that look like software faults
Hours get spent in configuration for problems that live in the first metre of the path.
Writing
Networking
Switching, addressing, packets and the faults that live between the cable and the application: how to find them, and how to design so they are rarer.
Three questions before touching a production network
Not process for its own sake. Three questions that have prevented more outages than any tool.
Jumbo frames: worth it, or not
They help in specific places and cause silent breakage everywhere else. Decide deliberately.
Broadcast domains and why they still matter
Switching is cheap, so people build large flat networks. Here is what that costs you.
When traceroute lies
Asterisks are not failures and high latency at hop four is usually nothing. Read it correctly.
TTL as a diagnostic instrument
The time-to-live field quietly tells you how many hops a packet took. That is more useful than it sounds.
Reading a packet capture without drowning
A capture with two million frames is not evidence until you have reduced it. Here is the order that works.
The difference between a slow network and a busy one
Both feel identical to users. They have different causes and different fixes.
Spanning tree: what it protects you from
A layer 2 loop does not degrade a network. It stops it, completely, in seconds.
ARP, and the day two devices claimed one address
Intermittent, symmetrical, affects everyone on the subnet. Duplicate addresses have a signature.