ΠΜΣ Πληροφορικής και Τηλεπικοινωνιών με ειδίκευση Τεχνολογία Συστημάτων Υπολογιστών
Library of the School of Science

2016-06-16

2016

Γάσπαρης Νίκων
Στρατής Νικόλαος

Δελής Αλέξιος

Αυτόματη υλοποίηση δυναμικού κώδικα για εκτέλεση πλάνων ερωτημάτων στη κυρία μνήμη με χρήση σύγχρονου υλικού υπολογιστών

Greek

Dynamic code generation for in-memory query plan execution using modern hardware

This postgraduate thesis studies whether when executing a join query,
generating real-time hand-written code results in better performance than using
an existing commercial database system.
We show that, for specific queries, just-in-time code generation is faster than
a conventional database management system. Note that when calculating the total
time needed to resolve a join query, we do not only take into account execution
time but also the time required for writing and compiling the query.
In this work we also examine two approaches for solving specific queries
(table-wise and variable-wise) and study how just-in-time code generation
affects their performance.
Specifically, we show to which extent the performance of each of the above
approaches is affected by the form of the query.
To evaluate this, we compare our system with a commercial database system using
both real data and random samples that were produced dynamically.

hand-written code, just in time code generation, join queries, variable-wise, table-wise

Yes

7,8,9

Yes

10

79

File access is restricted.

https://pergamos.lib.uoa.gr/uoa/dl/object/1316475