Thread: Re: [GSoC] application ideas
Hi, Michael! > 19 марта 2019 г., в 14:53, pantilimonov misha <pantlimon@yandex.ru> написал(а): > > 2) Changing buffer manager strategy. > Somewhere in 2016 Andres Freund made a presention[6] of possible improvements > that can be done in buffer manager. I find the idea of changing hashtable to > trees of radix trees[7] promising. Most likely, taking into account program's > time constraints, this task won't be done as "ready to deploy" solution. > Instead, some kind of prototype can be implemented and benchmarked. I like the idea of more efficient BufferTag->Page data structure. I'm not sure cache locality is a real problem there, butI believe this idea deserves giving it a shot. I'd happily review your proposal and co-mentor project, if it will be chosen for GSoC. Also, plz check some work of my students in related area [0]. Best regards, Andrey Borodin. [0] https://www.postgresql.org/message-id/flat/89A121E3-B593-4D65-98D9-BBC210B87268%40yandex-team.ru
Andrey, thank you for your reply. > 24.03.2019, 12:12, "Andrey Borodin" <x4mmm@yandex-team.ru>: > > I like the idea of more efficient BufferTag->Page data structure. I'm not sure cache locality is a real problem there,but I believe this idea deserves giving it a shot. > I'd happily review your proposal and co-mentor project, if it will be chosen for GSoC. Here it is: https://docs.google.com/document/d/1HmhOs07zE8Q1TX1pOdtjxHSjjAUjaO2tp9NSmry8muY/edit?usp=sharing > Also, plz check some work of my students in related area [0]. This is definitely helpful! Also found couple of relevant discussions, putting information together... -- Best regards, Michael.
Hi! We are discussing GSoC details offlist, but I'll put some recommendations on your proposal to the list. > 3 апр. 2019 г., в 2:53, pantilimonov misha <pantlimon@yandex.ru> написал(а): > > Andrey, thank you for your reply. > >> 24.03.2019, 12:12, "Andrey Borodin" <x4mmm@yandex-team.ru>: >> >> I like the idea of more efficient BufferTag->Page data structure. I'm not sure cache locality is a real problem there,but I believe this idea deserves giving it a shot. >> I'd happily review your proposal and co-mentor project, if it will be chosen for GSoC. > > Here it is: > > https://docs.google.com/document/d/1HmhOs07zE8Q1TX1pOdtjxHSjjAUjaO2tp9NSmry8muY/edit?usp=sharing While your project is clearly research-oriented, it is planned within PostgreSQL development process. Can you please addsome information about which patches are your going to put on commitfest? Also, please plan to review one or more patches. This is important for integrating into community. BTW, there is somewhat related IntegerSet data structure added recently [0]. In my version it was implemented as radix tree.I think it is a good example how generic data structure can be presented and then reused by BufferManager. Best regards, Andrey Borodin. [0] https://github.com/postgres/postgres/commit/df816f6ad532ad685a3897869a2e64d3a53fe312