CSE221 - lec06: Virtual Memory: VAX/VMS & Mach

06 - Virtual Memory: VAX/VMS & Mach

Virtual Memory Management in the VAX/VMS Operating System

Goals:

• support different hardware

• handle resource constraints (memory & I/O)
• Virtual Address Space: 4GB
• Physical Address Space: 250KB - 4MB

Paging Mechanisms

Space Efficiency

• process page table is paged (can be swap out)

• keep OS page table in physical memory

TLB

• stores va to pa mapping

• split between user and kernel TLB

Process-local replacement policy

• evict pages from the requested process only

• FIFO policy

Page Caching

• cache pages about to be evicted

• when page fault occures on pages that are in free list or modified list, they can be returned to resident set directly

Clustering

• read / write in batch & prefetch pages

Summary & take-away

• Address Space abstraction

• Hardware support for paging

• Techniques for efficiency

Machine-Independent Virtual Memory Management for Paged Uniprocessor and Multiprocessor Architectures

Goals:

• Machine independent approach to manage virtual memory

• hardware difference:
• single v.s. multi-processor
• non-uniform memory
• paging v.s. segmentation

• support sparse address space
• memory-mapped files
• multi-thread

Components of Mach VM Implementation: Address Map & Resident Page Table

Components of Mach VM Implementation: Memory Object

• backing store for a range of virtual address (resembles unix files)

• customizable pager

Components of Mach VM Implementation: pmap

• hardware-dependent module

• exact form differ wrt. hardware

Summary & take-away

• Mach VM’s approach to VM: build a machine independent virtual memory representation, and use a machine depedent module to do the translation.

• The idea is simple but the key is to keep the efficiency.