• who doesn't know mongo?
• motivation (big data)
• simple aggregation operations 
• aggregation framework
• map-reduce

• project - select, reshape - 1:1
• match - filter - n:1
• group - aggregate - n:1
• unwind - split - 1:n
• sort - 1:1
• skip - n:1
• limit - n:1
• others (redact, geoNear, out)

PIPELINE

db.collection.aggregate([ step_1, step_2, ..., step_n ])

• each step is one of 7 operations
  
• combining (repetative) operations in any order
• output of one becomes input of another 

• result set limited to 16MB of memory (the maximum BSON document size is 16MB as well)
• 10% of memory on a machine
• sharding

• 3 steps
• map function: scope
• reduce function: ~recursion, common key