parallelism_sadas_dbms

Optimization of the Intraquery parallelism (parallelism to optimize the performances of the single query) and Interquery parallelism (parallelism to optimize the performances of the engine with concurrent queries).

Action points:
• Improvement of the Heap management to reduce the negative effects of the Heap Contention (frequent concurrent requests of dynamic allocation of memory segments from threads).
• New algorithms for the automatic and/or guided management of the parallelism in the Nested Loops.
• Resizing of local buffers for the optimization of the use of the first and second level cache.
• Redesigning of the parallel algorithms and their scheduling to reduce the frequency of the switch context of the threads.

Results:
• Total throughput time of many concurrent queries improved by a factor of 10.
• Query response time improved by 50% with Union operators, and RAM maximum usage reduced by 50%.

 

Contact

(+1) 415 429 3969

20 California ST. 7TH Floor
SAN FRANCISCO, CA, 94111, USA

contact form

Login

This site uses cookies to improve the browsing experience of users

Continuing the navigation you accept the use of cookies Learn more

I understand

This site uses cookies to improve the browsing experience of users and to gather information about the site itself. You can know the details by consulting our privacy policy.
Continuing the navigation you accept the use of cookies; Otherwise you can leave the site.