CSE 544 Distributions

The Tukwila Execution Engine

See \\Data\Projects\Tukwila\544\Execution for QUERY.EXE.

The Tukwila execution engine runs on Windows NT machines only.  From the control window, you can open query plan files from disk and change the configuration settings.   It is highly recommended that you restart the execution engine after making configuration changes. 

Setting up the ODBC Wrapper for Your Data Sources

See \\Data\Projects\Tukwila\544\Wrappers for the .class files.

• Choose some database (e.g. Access, SQL Server, FoxPro, etc.)
• Go to the Windows Control Panel and create a System DSN (data source name) for the database application and the database within the application
• You'll need some version of the JDK, 1.1 or later.  Run one or more instances of the Windows NT Command Prompt in the appropriate directory, entering in each the command java ConsoleQuery plus (optionally) the following parameters
  • TCP port to listen on
  • Per-tuple delay to add, in ms
  • Initial latency to add, in ms
• Determine what your wrapper operator in Tukwila's plan will look like:

<N ID=id OP=wrapper TIMEOUT=timeInms CARD=expectedCard>
  <HOST>your machine hostname: your wrapper's port</HOST>
  <SERVICE>DSN name.database name</SERVICE>
  <QUERY>some SQL query</QUERY>

• Note that you may wish to run multiple instances of the wrapper, on different ports, for purposes of adding different wrapper delays.

The Data Source Generator

See \\Data\Projects\Tukwila\544\DataGen for the program and source.

The data source generator is not particularly sophisticated, but will take an SQL query or a comma-delimited text file and randomly sample some percentage of the file; it also allows the creation of disjoint peer-level data sources.

Instructions:

• From the NT Command Prompt, invoke with java DataGen dsn_name delimiter_string (you will need JDK 1.1).   By default, delimiter_string is a comma.
  
• Now enter your first definition following the syntax:

  outFile: ratio ( infile|"query" [- previousDefn [- previousDefn ... ]])

• For example:

  myout: 0.3 ( "select * from tpc.order" )-- myout  = 30% coverage of query
  otherout: 0.2 ( myout ) -- otherout = 20% coverage of myout

• You can also create disjoint data sets by using a difference operator.  Note that you can only do differencing between peer-level children of a common data source:

  mysecondout: 0.3 ("select * from tpc.order" - myout)
mythirdout: 0.3 ("select * from tpc.order" - myout - mysecondout)
  
  but not
  
  myfourthout: 0.2 ("select * from tpc.order" - otherout)

• Full Java source is provided that should be usable under any JDK 1.1-compliant environment.  Someone may wish to extend the program to do more than peer-level differencing (which will require tracing back to a common parent and determining tuple index equivalences), or to support saving of the sampled items between sessions.   Please share enhancements with your classmates.
• The output will be a series of tables (each with extension .TBL) with delimited values suitable for importing into a database.