Comparison of in-memory strategies, DBMS file in memory vs. Datablitz
The Lucent DataBlitz team loaded the white pages of a telephone book
onto a Sun computer system.
- One load was done with a commercial DBMS with the entire
database loaded into RAM.
- The other load was done under DataBlitz, with the entire
load in memory.
Update speeds, in updates per second, were as follows;
| Commercial DBMS |
560 |
| DataBlitz |
2,500-3,000 |
DataBlitz has a layered API. The layers
can provide a convenient API at the price of
slower speed, or very high performance
through a complex API. Its services can
be accessed from a relational layer, an intermediate
layer, and a low level layer at which locks and indexing
methods can be manipulated for maximum performance.
Lookup speeds, in lookups per second, were as follows:
| Commercial DBMS |
2,800 |
| DataBlitz/relational |
33,000 |
| DataBlitz/in-memory |
135,000 |