3. SMP Behaviour on Protein Sequence Data

Back to the start page

The following table shows the behavior of SMP on a 685MB Protein Sequence document for user-defined XPath queries. The following values are listed.

Proj. Size: Size of the projected document
Mem: High watermark of SMP memory consumption
Time: Total duration of document projection
User: User time spent for document projection
Sys: System time spent for document projection
Usr + Sys: Sum of User and Sys
CPU: Average CPU load during projection, computed as (User + Sys)/Time
States: Number of states of the constructed runtime automaton. CW (BM) denotes the number of Commentz-Walter (Boyer-Moore) automata constructed at runtime
Avg Jump Size: Average number of characters skipped while running CW/BM algorithm
Initial Jump: Percentage of document skipped by initial jumps
Char Comp: Total number of characters in the document that have been inspected

Projection benchmarks on 685MB Protein Sequence document

Runtime of SPEX for queries P1-P4 on 685MB document with and without projection

The projected documents are piped directly into the SPEX engine.

Throughput of SPEX for queries P1-P4 on 685MB document with and without projection

The projected documents are piped directly into the SPEX engine.

Queries

P1:	/ProteinDatabase//genetics[contains(mobile-element/text(),"retrovirus") or intron-status/text()="absent" or gene-origin/text()!="unknown"]/note
P2:	/ProteinDatabase/ProteinEntry[contains(descendant::group/text(),"Sao Paulo") and contains(sequence/text(),"LLLL") ]/header/uid
P3:	/*//reference/note[contains(child::text(),"Escherichia coli")]
P4:	/*//reference[contains(child::note/text(),"Escherichia coli")]//note

Back to the start page