How (not) to Size Your Servers
As a racing sailor I have been monitoring (and reporting on) the efforts by race organizers to track an ocean sailboat race and deliver live data on the web in map format. One such current event (Start Friday, June 16, First warning at 12:50hrs) is the biennial Newport to Bermuda Race.
This event in and of itself probably doesn’t merit a mention here. What does, in my opinion, is that after weeks of buildup to the “new satellite tracking system”, all a visitor gets four hours after the start is: “We are currently experiencing a very heavy load. Please check back later.”
Which begs the question: How was the system’s back end sized? Why was not the heavy load anticipated? And are similar surprises awaiting other such systems (AGX implementations, perhaps)?
Saturday morning update:
“Saturday, 0900: Server Overload
Due to everyone’s enthusiasm, the iBoattrack system is overwhelmed with requests.
Technicans are adding additional server capacity and hope the system will
be back online later on Saturday. We apologize for this server outage.
Below is a snapshot taken Saturday at 0400.”
This entry was contributed by Atanas Entchev
