I understand the swapping in Linux takes place not by a page by a page, several hundreds of pages in form of the clusters get swapped in and out at a moment. It would be great if I could get info regarding configuring the cluster size.
The idea of expelling pages in bunches is not new. It dates from the very early beginning of Berkeley UNIX and is documented in O. Babaoglu and W. Joy, "Converting a swap-based system to do paging in an architecture lacking page-reference bits," Proc. 8th ACM Symp. on Operating Systems Principles, pp. 78-86, 1981. I quote: "Having incurred the cost of page replacementpolicy activation, we would like to select more than a single page to be replaced in order to anticipate short-term demand for more memory. To this end, the system maintains a free page pool containing
all of the page frames that are currently not in the loop. Our version of the clock policy is triggered whenever the size of this pool drops below a threshold."
The size of the cluster in Linux while swapping is what I am referring to know about....Jehan-Francois Paris. It would be great if anyone would help me know that...:)