SimpleDB Overview

CSE 444 – Section 1

1

Today…

• Demo Git/Eclipse Setup
• Go through an overview of SimpleDB

2

3

Local Repository

/home/jortiz16

GitLab Individual Repository

git@.../simple-db-jortiz16

GitLab Course Repository

git@.../simple-db

Forked (by you)

Origin Remote

Cloned (by you)

Upstream Remote

Added (by you)

git push

git pull

git pull

upstream

master

Git

What you should NOT do:

• Modify given classes
• Removing, renaming, relocating to other packages
• Modify given methods
• Changing parameters or return types
• Use third-party libraries
• Except the ones under lib/directory
• You can do everything using regular Java libraries

4

What you CAN do:

• Add new classes/interfaces/methods/packages
• Watch out for name conflicts with future labs!
• Safer choice: use new packages (best) or inner classes (meh)
• Re-implement provided methods
• Just don’t destroy correctness or specification!
• Find bugs!

5

What you CAN do (continued):

• System test cases
• Under test/systemtest
• We’ll grade using additional tests
• Write up
• Explain why do you implement in that way
• We’ll read your code
• Reading horrible code is horrible, so spend some time polishing
• Passing all the test cases may not necessary mean you’ll get a high score
• Sanity check, not a final grade

6

Setting up SimpleDB

Any questions or concerns?

7

Overview of SimpleDB

8

Database

• A single database
• One schema
• List of tables
• References to major components
• Global instance of Catalog
• Global instance of BufferPool

9

Catalog

• Stores metadata about tables in the database
• void addTable(DbFile d, TupleDesc d)
• DbFile getTable(int tableid)
• TupleDesc getTupleDesc(int tableid)
• …
• NOT persisted to disk
• Catalog info is reloaded every time SimpleDB starts up

10

BufferPool

• The ONLY bridge between data-processing�operators and actual data files
• Strict interface for physical independence!
• Data files are never accessed directly
• Later labs:
• Locking for transactions
• Flushing pages for recovery

11

Data Types

• Integer
• Type.INT_TYPE
• 4 byte width
• Fixed-length Strings
• Type.STRING_TYPE
• 128 bytes long (Type.STRING_LEN)
• Do not change this constant!

12

OpIterator

• Ancestor class for all operators
• Join, Project, SeqScan, etc…
• Each operator has methods:
• open(), close(), getTupleDesc(), hasNext(), next(), rewind()
• Iterator model: Chain iterators together

13

HeapFile

• Main class that organizes the physical storage of tables
• Collection of HeapPages on disk
• One HeapFile for each table
• Fixed-size pages means efficient lookup of pages

14

15

HeapPage

• A chunk of data that can reside in the BufferPool
• Format: Header + Tuples
• # of 1 bits in Bitmap = # of active tuples on page
• Fixed size: BufferPool.PAGE_SIZE

16

Questions?

17