Survey: Computational Models for Asymmetric Read and Write Costs

With the arrival of new emerging technologies for non-volatile main memory, the cost of reading from memory will be significantly cheaper than the cost of writing to memory. This raises new challenges to algorithm design since many of the classic algorithms use approximately the same number of reads...

Celý popis

Uloženo v:

Podrobná bibliografie
Vydáno v:	2018 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW) s. 733 - 743
Hlavní autor:	Gu, Yan
Médium:	Konferenční příspěvek
Jazyk:	angličtina
Vydáno:	IEEE 01.05.2018
Témata:	Analytical models asymmetric memory Complexity theory computational model Computational modeling I/O efficient algorithms Memory management nested parallel computation non volatile memory Phase change random access memory Task analysis write efficient algorithms
On-line přístup:	Získat plný text
Tagy:	Přidat tag Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!

Popis
Shrnutí:	With the arrival of new emerging technologies for non-volatile main memory, the cost of reading from memory will be significantly cheaper than the cost of writing to memory. This raises new challenges to algorithm design since many of the classic algorithms use approximately the same number of reads and writes. In this survey, we review the existing computational models that measure the cost of operations and memory accesses to the NVMs, and scheduler of parallel algorithms on the new hardware. The new models meanwhile allow theoretical analysis and showing lower bounds on the hardness of the classic problems. We also list some existing results on lower and upper bounds on the most common problems like sorting, searching, graph traversal, based on these models.
DOI:	10.1109/IPDPSW.2018.00120