Question

Suppose that a huge file is stored in secondary storage. It is much too large to fit into main memory. Describe an algorithm for sorting the file. You should try to minimize the number of page reads and writes. You can create extra files, of any size, on secondary storage if you need them. Say what you can about both the run time and the number of page accesses used by your algorithm. To make things definite, you might consider the case where the file is 1 TB (240 bytes) and main memory is 8 GB (233 bytes) and page size is 4 KB (212 bytes). (You can assume, if you like, that no item in the file overlaps a page boundary. That is, when you read a page, you get some whole number of items.)

Answer 1

We have a file of size 1TB

We will divide the file in chunks of 8GB. so total there will be 2^40/2^33 = 2^7 = 128 chunks

To get 8GB of data from disk we may have to use pages so total 2^33/2^12 = 2^21 pages will be used.

We will sort 8GB of data using mergesort. and store it back on disks

We can create a minheap of size 128 and read one element from each chunk at a time. Initially we will insert the first element that is the smallest element in chunk in heap.

When we extract the element we will get the smallest element among and insert the element from that chunk from which the element is extracted.

doing this for 2^33 times we will get a sorted 1TB file.