Transcript Virtual Machine Monitors

Virtual Machine Monitors
CSE451
Andrew Whitaker
Hardware Virtualization
Running multiple operating systems on a
single physical machine
Examples:
VMWare, Microsoft’s VirtualPC / VirtualServer,
Parallels (Macintosh), Xen
Virtual Machine Monitors
Type I VMM runs on the raw hardware
Type II VMM runs hosted on another OS
Windows
Linux
virtual machine
virtual machine
virtual machine monitor
hardware
VMM History
Conceived by IBM in the late 1960’s
Popularized by VM/370 (1972)
Used for OS debugging, time sharing,
supporting multiple OS’s
Batch processing
Time sharing
OS
OS
VM/370
System 370 Machine
VMMs Today
OS development and debugging
Software compatibility testing
Running software from another OS
Or, OS version
Virtual infrastructure for Internet services
Internet Services
 Many applications now run inside the Internet
Search engines, maps, photo sharing, email, IM, game
servers, etc.
 Goal: allow anyone to upload to a new service
into the Internet
Must be low-cost, secure, robust, easy to maintain, etc.
 Approach: use VMMs to provide a rent-a-server
economy
Amazon’s Elastic Compute Cloud
(EC2)
 Provide service developers with a set virtual
machines and storage resources
 Scalability
New machines can be created in minutes
 Security
Virtual machines provide stronger isolation than OS
processes
 Developer control
Developers choose the OS, software, libraries, etc.
 Low cost
Developers pay only for what they use
VMM Implementation Overview
A VMM is just an operating system that
exposes a (virtual) hardware interface
Windows
Linux
virtual machine
virtual machine
virtual machine monitor
hardware
virtual
architecture
=
physical
architecture
Comparing the Unix and VMM APIs
UNIX
VMM
Storage
File system
(virtual) disk
Networking
Sockets
(virtual) Ethernet
Memory
Virtual Memory
(virtual) Physical
memory
Display
/dev/console
(virtual) Keyboard,
display device
Possible Implementation Strategy:
Complete machine emulation
The VMM implements the
complete hardware architecture
in software
while(true) {
Instruction instr = fetch();
// emulate behavior in software
instr.emulate();
}
Drawback: This is really slow
Review: Protection in Traditional
OS’s
 Hardware exposes a user/kernel boundary
Operating system runs in kernel mode
Processes run in user mode
 Processes can safely execute most instructions
No need to “emulate” machine behavior
 But, a small set of instructions can only execute
in privileged mode
e.g., writing to an I/O device, disabling interrupts, etc.
Improving VMM Performance
 Treat guest operating systems like an application
 Most instructions execute natively on the CPU
 Privileged instructions must be trapped and emulated
Virtual machines
loads,stores,
branches,
ALU operations
VMM
Physical hardware
machine halt,
I/O instructions,
MMU manipulation
Handling Privileged Instruction
 Virtual machine issues a privileged instruction
(e.g., disk read)
 VMM determines whether the virtual machine
was in “user” mode or “kernel” mode
Note: the virtual mode is distinct from the physical mode
(Yikes!)
 If “user” mode, raise a protection exception
 If “kernel” mode, emulate the disk read in
software; then, return control to the guest OS
Tracing Through a File System Read
Application
Guest OS
VMM
read() syscall
trap handler
handle read syscall
read from disk()
finish read syscall
copy data to user buffer
return from system call
return from read()
trap privileged instruction
If “kernel” mode:
emulate virtual disk
else:
raise protection violation
Virtual Disk: Possible
Implementations
Static disk partitions
A file in the file system
Especially for type-II VMMs
A special virtual disk file system
A network storage abstraction
e.g., Amazon’s S3
Virtualizing the User/Kernel Boundary
 Both the guest OS and applications run in
(physical) user-mode
This is necessary so that privileged instructions trap into
the VMM
 For each virtual machine, the VMM keeps a
software mode bit:
During a system call, switch to “kernel” mode
On system call return, switch to “user” mode
 Note: A faster implementation is possible on x86
(Virtual) Physical Memory
Each guest OS expects its own page
tables, TLB, memory-management unit,
etc.
Unlike the virtual disk, we can’t trap and
emulate every memory reference
WAY too slow
Additional complication: protection bits
Some memory can only be accessed in “kernel”
mode
Memory Model
Virtual Memory
(Applications)
Increasing
privilege
Physical Memory
(Guest OS)
Machine Memory
(VMM)
Assume a software-loaded TLB:
InsertTLB(int virtualPageNumber,int physPageNumber);
Virtual Physical Memory,
Implementation
(Bad) Option #1: Insert the V-to-P mapping
directly into the TLB; Trap each memory
access to return the V-to-M memory
Better Option #2: Instead of inserting a Vto-P mapping, transparently insert a V-toM mapping
This way, all future memory references will
occur at hardware speed
Performance Details
 TLB must be flushed on a process switch and a
virtual machine switch
 Whenever a virtual machine regains the
processor, it requires numerous page faults to
restore its virtual memory mappings
 Optimization: VMM maintains a cache of V-to-M
mappings for each virtual machine
On taking a page fault, the VMM can insert the
appropriate TLB mapping without consulting the guest
OS