HotNets Slides

• talk about latchups first to no longer need to talk about heat generation

• bring up solar panel power issue

• point out latchup physically and talk about how it’s noisy

• graph elaboration

  • x axis
  • y axis
  • slide titles
  • show key takeaways of graphs on the slides

• improve motivation: Solar Superstorms

  • look at video on ACM DL (short video)

• fix flat intonation

• Pauses between slide still gotta be longer

• At the end. “ready to answer questions”

• Answering questions

  • Repeat questions being asked
  • More than two round-trip questions: “let’s discuss this offline”
  • Add GitHub links?

• conclusion

  • future work
  • cool ideas to work on in the future

• add an outline slide about what we’re talking about

• emphasize looking at the main ideas

• elaborate on terms for control-flow integrity

• key insight, change pitch: emphasize the black-box thing more

  • “we’ll talk about concrete examples later”

• make the opening better introduce urself and work briefly

  • this is a joint work between…

• focus one idea for each SEL/SEU

  • drop out the ECC parts, add a “read our paper for details” on other systems
  • dependencies between data/control flow -> catch the subset
  • just primarily about the control flow

• explain anything being shown in the slides

  • especially for graphs, explain x and y axes and what lines are

• use a single graph for SEL explanation and explain more about the graph

• Pause for ~3 seconds during each slide

• Change titles to highlight key points

• Explain why heat and energy generation is bad

• Cut down on floating-point stuff (reduce specificity of approach sections)

• Be more assertive -> a protects b (don’t convey doubt with “should be” etc.)

• Emphasize key insight of prior work treating as black-box earlier right after

  • Solutions are just examples

• DSP: talk about how most applications/flight software don’t use the DSP functions

• Talk about curve with data about increase percentage

• ask SNL for feedback