List of terms specific to the thesis
This section lists a) terms specific to this thesis and b) terms used with a specific meaning. Compound terms based on these, and incorporating the concepts of average, minimum, maximum and difference (e.g. avgQS, minAS, ΔSTn,m) are not included here, but are explained in full where they first occur.
Appendix 6 is a glossary of general acronyms and terms.
list of specific acronyms
- AS
- access size.
- NPA
- number of partitions accessed used by a range query.
- migprob
- migration probability.
- pcost
- partitioning cost.
- PS
- partition size.
- QS
- query size.
- STS
- scan time saving.
- WL
- workload.
- WLS
- workload saving.
- WW
- wasted work.
glossary of terms
- access size
- the logical fraction of data accessed in a given dimension.
- end effect
- the breakdown in the assumption that range queries evenly reference the domain range. This breakdown occurs when the number of partitions is small.
- migration cost
- the cost of a single data migration between two partitions.
- migration probability
- the probability that an OLTP transaction will cause an update to a record which will require that record to migrate.
- partitioning cost
- the overall cost of partitioning a column.
- query size
- the fraction of records referenced by a range query in a given dimension.
- scan time saving
- defined as the cost of scanning the partitions in which required data resides minus the cost of scanning the data if there were no partitioning.
- wasted work
- amount of data accessed minus the required data. In a single dimension this is QS-AS.
- workload
- defined as the scan time plus partition costs.
- workload saving
- the total saving in workload that can be made by partitioning a file (= STS + pcost).