W4121: Non-graded Quiz, Wu’s Lectures

Click Here For Solutions

Lec2: Data Modeling

The point of data modeling is to think about what characteristics of the data you will need for your application.

Q1: Data Independence

1. What is data independence and why does it matter?
2. What are the main two reasons why hierarchical data model is a bad idea?
3. Give an example that illustrates the bad idea.
4. When could it make sense to use a hierarchical data model?
5. What did the graph model try to solve? What limitations did it still have?

Q2: Queries

Suppose our data adheres to the following schema

    animals(id, species, name, gender, keeper_id)
    keepers(id, name, birthday)
    feedings(keeper_id, animal_id, date, time, food, pounds)

Describe in english what each of the following operations will do:

• filter(feedings, time < 12pm)
• filter(feedings, time < 12pm and pounds > 10)
• project(animals, [name])
• project(animals, [species])
• join(feedings, keepers, keeper_id = id)
• join(animals, feedings, keeper_id = id)
• groupby(feedings, [food], SUM(pounds), COUNT(pounds))

What are the operator symbols (greek symbols) for the above operations?

Describe in english what the following operations do:

• Π_food σ_time=12pm(feedings)
• Π_name (σ_time=12pm(feedings) ⋈_{id=animal_id} σ_{species=gorilla}(animals))

Q3: Logging

We have seen that having a schema is helpful because then data can be easily queried. Before the solution described in the paper, Twitter’s event messages DID adhere to a schema of (category, message).

1. Even though the events had a schema, why did data scientists still have trouble analyzing logging data?
   • What was the source of this trouble?
   • Why do you think Twitter let this happen? Couldn’t they have prevented it from the very beginning?
2. What was the main solution to address the above troubles?

Lec 3/4: Query Execution