Notes from Startup School - "Random Stuff" by Paul Buchheit

Paul Buchheit (blog) - Creator of GMail

Paul Buchheit speaking at Startup School.

Advice on advice

• where does advice come from -- limited life experiences + overgeneralization = advice
  
• doesn't mean that advice is worthless, have to remember where it is coming from.
  
• helpful advice
  
  • opens your eyes so you can see more clearly,
    
  • gives close about the overall picuture, how your part fits in
  
  
• watch out advice that is limiting. Remember that no one hs it all figured out.
  

Redefine success.

Redefine success

• attempt things with uncertain outcomes - but uncertain outcomes that you'll learn from
  
• redefine goal of startup - if learning is goal, then startups are a great place to learn, and you can achieve that goal
  

Make something that is different

• how can you tell hat a product is fundamentally different? by how people use it -- Altavista, 2 times a week. Google, 10 times a day.
  
• please remember - being useless in new ways does not make you different from previous useless products
  

GMail - changing the magnitude.

Quick tips on scaling web service

• disks are not really random access - capacity doubles every year or two, but seek times remain the same
  
• relative seek time is actually decreasing, as disks get bigger.
  
• solution - think of disk drive as a sequential access device
  
• how build around this?
  
  • big data, small data
    
  • big - photos, video, mp3 (you can just put it in amazon s3)
    
  • small - tags, names, meta data (keep it in RAM)
    
  
  
• DB updates must avoid cause seeks - updates go to a transaction log. All writes are sequential.