H ow the ENVS and QRES stacks work (example)

5
K.Subieta. How the stacks work, Slide 1 3rd Country Conference, Feb.200 How the ENVS and QRES stacks work (example) by Prof. Kazimierz Subieta [email protected] http://www.ipipan.waw.pl/~subieta SBA/SBQL pages: http:// www.sbql.pl 3rd Country Conference, February 2007

description

H ow the ENVS and QRES stacks work (example). 3rd Country Conference, February 2007. by Prof. Kazimierz Subieta subieta@pjwstk. edu .pl http://www.ipipan.waw.pl/~subieta SBA/SBQL pages: http://www.sbql.pl. Query addressing a tiny database. - PowerPoint PPT Presentation

Transcript of H ow the ENVS and QRES stacks work (example)

Page 1: H ow the ENVS and QRES stacks work (example)

K.Subieta. How the stacks work, Slide 1 3rd Country Conference, Feb.2007

How the ENVS and QRES stacks work (example)

byProf. Kazimierz Subieta

[email protected]://www.ipipan.waw.pl/~subieta

SBA/SBQL pages: http://www.sbql.pl

3rd Country Conference, February 2007

Page 2: H ow the ENVS and QRES stacks work (example)

K.Subieta. How the stacks work, Slide 2 3rd Country Conference, Feb.2007

Query addressing a tiny database

• Query: For each department get its reference and the average sal of its employees.

• SBQL: Dept join avg(employs.Emp.sal)

• Tiny database:

• Expected result: bag{ struct{i17, 2500}, struct{i22, 1450} }

Page 3: H ow the ENVS and QRES stacks work (example)

K.Subieta. How the stacks work, Slide 3 3rd Country Conference, Feb.2007

Execution scenarioDept join avg( ( employs . Emp ) . sal )

join 1st dot 2nd dot avg

1st Dept

2nd Dept

1st employs

2nd employs

1st Emp

2nd Emp

3rd Emp3rd employs

Page 4: H ow the ENVS and QRES stacks work (example)

K.Subieta. How the stacks work, Slide 4 3rd Country Conference, Feb.2007

Sample queryDept join avg( ( employs . Emp ) . sal )

Emp(i1) Emp(i5) Emp(i9) Dept(i17) Dept(i22)

ENVS QRES

empty

Start

Page 5: H ow the ENVS and QRES stacks work (example)

K.Subieta. How the stacks work, Slide 5 3rd Country Conference, Feb.2007

In closing …

• Please note that this evaluation scenario concerns the formal semantics only, executed by an abstract machine.– Our goal is clear, universal and free of ambiguities semantic

specification rather than performance.

• In real implementation there are many obvious optimizations.– C.f. our implementations in various projects.

• The evaluation process is also the subject of many sophisticated optimizations (rewriting, indices, caching,…) .

• The entire process is executed in main memory, hence the computations are not critical.– Critical computations concern the access to external storage and

network, and avoiding unnecessary (redundant) computations.

– This is just the subject of major SBA optimization methods.