The Whats and Whys of Whole System Virtualization

1
The Whats and Whys ofWhole System Virtualization

• Peter A. Dinda
• Prescience Lab
• Department of Computer Science
• Northwestern University
• http//plab.cs.northwestern.edu
• Virtuoso Project http//virtuoso.cs.northwestern.
  edu

2
Whole System Virtualization

• Many problems in computer science can be solved
  by adding a layer of indirection (bad
  paraphrase)
• Virtualized X
• X is a semantically invisible layer of the
  software stack
• Exports exactly the interface it builds on
• Adds functionality and/or solves problems
• Whole system span the stack horizontally

3
OS Virtual Machines

• Traditional (Goldberg types I and II)
• Run off-the-shelf operating systems
• Very low computational overhead but some I/O
  overheads (arguable how far it can be reduced)
• VMware, Microsoft (and VM from the early 70s)
• Paravirtualized
• OS kernels must be ported to them
• Very low computational and I/O overhead
• Xen, User Mode Linux
• Virtual servers
• OS kernel extensions (one OS, many instances)
• Negligible computational and I/O overhead
• Vserver, BSD Jails

4
Language Virtual Machines

• Compiler targets abstract machine
• Usually stack machine
• Run-time interprets and dynamically translates to
  base ISA
• Large standard library for I/O
• JVM, CLR, (and p-System from late 70s)
• Arguably also Lisp, Scheme, Perl, Python,

5
Overlay Networks and P2P

• VPNs and VLANs
• Multisource multicast (ESM, etc)
• Distributed hash tables (Chord, etc)
• Resilient routing (RON, etc)
• Anonymous routing (Tor, etc)
• VM-specialized (VNET, VIOLIN)

6
Virtual Storage And Devices

• Storage Area Networks
• iSCSI
• Remote device support
• Network block device

7
Virtualized Services

• Tunneling
• ssh
• Virtual file systems
• System-call interposition

8
Reducing Complexity

• Ownership
• Give the user the parallel/distributed systems
  analogue of a PC
• Deployment and distribution
• Whole system image
• See Potters snapshots for a very nice example
• Automatic policy avoidanceWnavigation
• Route through the diverse security policies in a
  multi-site computing environment

9
Adaptive Systems
VM Layer

• Bring automatic adaptation and resource
  reservations to existing, unmodified
  applications
• Virtualization as a layer for observation, a
  provider of adaptation mechanisms, and an
  impedance matcher to reservations

Virtualization Layer
Physical Layer
10
Making High-end Computing A Commodity

• Virtualization for fungibility
• Providers perspective
• Simple, straightforward abstraction to sell
• Users perspective
• Maximum flexibility
• Giant PC

11
Open-source Virtual Machine Monitor

• Type-I OS VMM for modern architectures
• Intels VT extension to IA32 and IA32e, and AMDs
  Pacifica extension to AMD64
• Make these commodity architectures virtualizable
  in the Goldberg sense
• VT/Pacifica VMM can be MUCH simpler than existing
  VMMs for these architectures
• Think 50K lines of code (VAX Secure VMM example)
• Potentially a very high impact project from this
  community

12
Trustless Computing and Language VMs

• Trust asymmetry problem in grid and utility
  computing
• Encrypted computation to the rescue
• Language VMs are perfect place to implement
• Translate binary to binary
• Portable