Dynamic Analysis And Profiling Of Multi Threaded Systems
Memory Management In Vmware Esx Server
1. Summary of:- Memory Management in Vmware ESX
Server
Carl A. Waldspurger
Vmware, Inc.
Submitted by:- Ubaidullah Alias Kashif
Submitted to:- Dr Qamar-ud-din Khand
Subject:- Advance Operating System
Class:- MSCCN-III
Sukkur Institute of Business Administration,
airport raod Sukkur
2. SUMMARY
Virtualization is very essentail as it is observerd from current market
scenorio. If we talk about underutilization of individual servers we can consolidate
servers into big one and letting each of which behaves seperatly or a big consolidated
servers that have multiprocessing capability can be used as few independent machine
performing different task. That is because of virtualization techniques. ESX Sever is a
software layer at which hardware resources are multiplaxed efficiently, it provides
higher degree of I/O performance such that is directly manages system hardware
with out interception of VMM. ESX Sever gives each VM an illusion of having zero-
based physical address as proviede by real hardware. It has dynamic way for address
of VMs means if address size exceeds than configured size it ll automatically
manage. In memory overcommiting there is issue of page replacement as a solution
it intoduces another level of paging known as meta-level page replacement where for
reclaiming VM’s particular pages are also reclaimed.ESX sever loads a module into
guest OS called ballon as a kernel service. In order to reclaim memory “ in-flate”
operation is performed and pinned physical pages are allocated within VM. By “de-
flate” operation memory is deallocated.Data that is commmon and is shared among
different applications running on same or different guest OS. Transparent page
sharing is a method in which redundent copies are eliminated. Common apges are
indetified and mapped to same machine page, and marked copy-on-write. Content
based page sharing is mechanism in which identical contents are identified and
shared regardless where and how those contents were generated. In order to make
content based page sharing efficient we can compare content of each page that is
much coslty, we can use hashing techniques to match the page. Resource allocation
share based is not an efficient due to underutilization of resources thae are shared so
for this randomized and determistic algorithm have been proposed such as min-
funding revocation algorithm. Page remapping yields reduced I/O copying
overheads.