[Spark MOOC note] Lec6. Structured Data

RELATIONAL DATABASE

review: key data management concepts:

  • data model
  • schema
  • relational data model

structured data: have a specific schema to start with

relationl database: a set of relations. 2 parts to a Relation:

  • schema: name of relation, name and type of columns

  • instance:

any data at given time (cardinality:=nb of rows, degree:=nb of fields)

LARGE DATABASES

RELATIONAL DATABASE EXAMPLE AND DISCUSSION


cardinality=3 degree=5

advantages of Relational Databases:

  • well-def structure
  • maintain indices for high performance
  • consistancy maintained by transactions

disadvantages:

  • limited, rigid structure
  • most disk space taken by large indices
  • transactions are slow
  • poor support for sparse data(which is common)

STRUCTURED QUERY LANGUAGE (SQL)

supported by DataFrame of pyspark

JOINS IN SQL

cross join: carteian product

EXPLICIT SQL JOINS

explicit version is preferred

TYPES OF SQL JOINS

⇒ controls how unmatched keys are handled

LEFT OUTER JOIN: keys appearring in left table but not in right table will be included with NULL as value

JOINS IN SPARK

  • for spark DataFrame: support inner/left outer/semi-join
  • for pair RDDs: support inner join(), leftOuterJoin(), fullOuterJoin()

join ex:

outerjoin ex:

fullouterjoin ex:

Lab 2 - Web Server Log Analysis with Apache Spark

Apache Common Log Format (CLF):
127.0.0.1 - - [01/Aug/1995:00:00:01 -0400] "GET /images/launch-logo.gif HTTP/1.0" 200 1839

Row( host = match.group(1), client_identd = match.group(2), user_id = match.group(3), date_time = parse_apache_time(match.group(4)), method = match.group(5), endpoint = match.group(6), protocol = match.group(7), response_code = int(match.group(8)), content_size = size )

  • distinctByKey

一个pair RDD按照key来distinct不知道有没有distinctByKey之类的东西, 只好写成这样, 不知是不是对的: dayHostCount = dayGroupedHosts.map(lambda group : (group[0], len(set(group[1])) ) )

...总体来说很有意思的一个lab...

comments powered by Disqus