High Availability of Virtualized Desktop Applications

Document Sample
High Availability of Virtualized Desktop Applications Powered By Docstoc
					High Availability of
Virtualized Desktop
    Applications
      Todd Deshane
  CS Seminar Fall 2006
Advisor: Jeanna Matthews
               Overview
 Why High Availability?
 Why Virtualized?

 Why Desktop Applications?



 Implementation   Sketch
 Discussion

 Future   Work
     Why High Availability?
 Malware   (viruses, worms, spyware,
  etc.) are still a problem
 Patches and software bugs can cause
  problems
 Fixing or reinstalling is time
  consuming and difficult for most
  computer users
 Application downtime means
  decreased user productivity
             Virtualization

 Level of abstraction
 Which types of virtualization are of
  interest to us?
  – Architecture Virtualization (Virtualization
    of underlying hardware i.e. x86)
  – OS Virtualization (virtualization of
    underlying kernel)
Architecture Virtualization

            Guest Operating Systems “think”
            they have full access to the
            underlying hardware



            Virtual Machine Monitor (VMM)
            provides an interface to the
            underlying hardware
OS Virtualization
       Guest Operating Systems (Virtual
       Private Servers) “think” they have
       full access to the underlying kernel




       Virtualization Layer in kernel
       provides interface to kernel
           Why Virtualized?
 More   effective use of hardware
  – Desktop utilization less than 10%
 Isolation   of guests
  – Guests run different applications
 Recovery    of compromised guests
  – Restarting or rolling back a guest is
    easier than restarting or rolling back a
    physical computer
   Why Desktop Applications?
 Desktop users’ increased reliance on
  power of computers
 Increased sophistication of attacks
  – Distributed botnets
  – Criminal attacks on data and resources
               and HA cluster
 Virtualization
 research traditionally focused on
 servers
      Implementation Sketch
 Rapid   Recovery System
  – Isolated virtual machine appliances
  – Intrusion detection and restoration
 High   Availability Cluster
  – Redundancy of applications
  – Heartbeat monitoring
 Application   Service Provider
  – Central application server
  – On-demand software
Rapid Recovery System
       Rapid Recovery System
 Key    Features
  – Network Intrusion Detection
  – Data Protection
  – Fast Restoration
  – File System Intrusion Detection
        Read   and write limiting
 To   be added:
  – Contract-based data access
High Availability (HA) Cluster




With a passive stand-by
system, fail-over can happen
almost instantaneously
High Availability (HA) Cluster




With a shared disk, data recovery
can happen almost instantaneously
 High Availability (HA) Cluster
 Key   Features
  – Heartbeat Monitoring
  – System or Data Fail-over
  – Reliability
  – Resource Management
     Data redundancy
     System redundancy

     Configurable Policies
Application Service Provider (ASP)
                  Application server
                  serves requests for
                  applications




                     Thin clients request
                     applications for server
Application Service Provider (ASP)
 Key   Features
  – Central Application Server
  – Single backend storage connection point
  – Stateless Configuration Possible
   Implementation Integration
 ASPs  are put into virtual machine
  appliances
 Rapid recovery system protects user
  data and settings and restores ASPs
 High availability cluster provides
  redundancy of application servers
  and heartbeat monitoring
              Discussion
 Users increasingly rely on desktop
  applications
 Attack sophistication is increasing
 Combine virtualization, HA cluster,
  and rapid recovery technologies
 High availability of virtualized
  desktop applications
  – Attack resistant desktop experience
  – Rapid application and data recovery
                Discussion
 Application   deployment model
  – ASPs can be local or remote
  – User chooses which ASP version to use
 Software   Redundancy
  – Most faults occur in software
 Systemwill also support hardware
 redundancy (if available)
           Future Work
 Deployment   and testing of HA cluster
  technology
 Virtualization performance and usage
  testing on the desktop
 Enhance and test the implementation
  of the rapid recovery system
 System integration of these existing
  technologies
Questions?
Backup Slides
         Implementation Details
 Virtual   Environment:
  – Xen (VMM)
  – OpenVZ (OS virtualization)
 High    Availability Cluster
  – Linux-HA
 File   Server:
  – Openfiler
             References
 Matthews    et al. Data Protection
  and Rapid Recovery From Attack
  With A Virtual Private File Server and
  Virtual Machine Appliances
 www.linux-ha.org/

 http://www.openfiler.com/

 http://en.wikipedia.org/wiki/Applicati
  on_service_provider