Hello all
We are developping a software that has a lot of concurrent transactions so we use a lot row locking (SELECT.. FOR SHARE/UPDATE) and we are experiencing high disk write rates on large read queries.
As I understand the tuple is updated every time a lock is put on it (then page becomes dirty and at some point is written to the disk)
Is there any way to prevent the writing to the disk since this information is temporary? Is there any system for caching the rows locks in RAM in PostgreSQL? Is increasing shared_buffers a solution since it contains the cached rows or is it only a read-purpose caching?
Thanks in advance for your knowledge on the subject
Have a nice day
Martin