One of the major reasons for the popularity of OLAP servers has been their speed in resolving queries. Numerically speaking, they are capable of producing an answer to even the most complex queries in approximately 0.1% of the time as would be taken on OLTP relational data, which is basically the concept upon which OLAP has been developed, so it obviously makes sense for it to be faster than the other. But the main reason for such phenomenally high speed is the use of aggregations, that are built from the fact table following a change in the granularity on specific dimensions and aggregation of data along these very dimensions. The number of aggregations that are possible in each case depends on every possible combination of dimension granularities.
Further, since there are numerous aggregations that can be calculated on OLAP servers, it often happens that only a specific number of the same are fully calculated, with the rest being solved on demand. The problem of choosing which of these aggregations need to be calculated is commonly known as the 'View Selection' problem, which can be restricted by the size of the selected set of aggregations, time to update them from changes in the base data, and in some cases, both. The basic purpose of view selection is to reduce the average time to respond to OLAP queries, so as to further enhance the utility of OLAP servers for organizations.
No comments:
Post a Comment