Visualizing these numbers is a good idea, but this is a terrible execution.
- The line segments go in different directions, making it unnecessarily difficult to visually determine relative lengths
- The packet send from CA to the Netherlands and back to CA is split over many lines, making it even more difficult
- The data is split across multiple maps, making it nearly impossible to get a good sense of, say, just how much longer the packet roundtrip takes than a memory reference.
- It's nearly impossible to see the line segment for main memory reference
- Having the lines overlaid on a map is more distracting than anything else
That was an awesome lecture. Thanks for the link. I did not know Hopper was such a humorous lecturer. The audience wasn't so lively - I guess they were mostly military.
I thought that the whole point of converting these latency numbers was to convert them into something meaningful. Distance traveled by the speed of light is just as abstract as the original number itself.
6 comments
[ 4.2 ms ] story [ 22.6 ms ] thread- The line segments go in different directions, making it unnecessarily difficult to visually determine relative lengths
- The packet send from CA to the Netherlands and back to CA is split over many lines, making it even more difficult
- The data is split across multiple maps, making it nearly impossible to get a good sense of, say, just how much longer the packet roundtrip takes than a memory reference.
- It's nearly impossible to see the line segment for main memory reference
- Having the lines overlaid on a map is more distracting than anything else
A good metric to compare those numbers against.
[1]http://www.stuartcheshire.org/papers/LatencyQuest.html