Overview

This course was taught at MIT during its Independent Activities Period in January 2010 by employees of VMware, nearly all of whom have close ties with MIT, and are excited by the intellectual challenges arising from the technology. The course consists of 5 days of material that cover the essentials of virtualization technologies. Although each day’s lectures are self-contained, they are meant to cover the spectrum: CPU, network, and storage virtualization, in both the large and small, e.g. cloud and mobile. Each topic will be motivated by a particular need and the various solutions will be compared.

Lectures

• Introduction to Virtual Machines (Carl Waldspurger, Kenn Bar, & Ravi Soundararajan)
  

    Introduction to Virtual Machines

    Introduction to Virtual Machines

  Introduction to Virtual Machines (Video)
  Many Macintosh users are familiar with virtual machine technology that enables Microsoft Windows or Linux operating systems to run on their machines, some PC users have experience with similar technology, and many data center managers know the economic benefits of running virtual machines within their servers. Experience using these products is different from understanding how they work.Isolation, encapsulation, and interpositioning are three benefits of virtualized machine technology. This lecture will cover some of the techniques used to virtualize X-86 processors in general, and the CPU, memory, and peripheral devices in particular. Binary translation and hardware-assisted virtualization will be explained.

  Virtualization, by its very nature, incurs a performance overhead, however, it is minimal. How minimal? The presentation will include performance analysis of some of these overheads.

• Virtualization of Networks and Storage in the Data Center (Yong Wang & Larry Rudolph)
  

    Network Virtualization

    Network Virtualization

  Network Virtualization (Video)

    Storage Virtualization

    Storage Virtualization

  Storage Virtualization (Video)
  The modern data center has three complex components: servers, network, and storage. Virtual machines are mapped to the servers but can be migrated between servers to optimize the overall performance.Similarly, virtual disks can be mapped and migrated among storage devices. The fun happens when the physical network and storage devices provide services above the basic functional operations. For example, provisioning different bandwidth limits in the physical domain can adversely affect the mapping and migration of the virtual components.

  There will be two lectures covering network and storage. Each will motivate the challenges by considering advanced services provided by networks and storage devices that usually are not discussed in core undergraduate CS courses.

• Resource Management (Carl Waldspurger)
  

    Resource Management for Virtualized Systems

    Resource Management for Virtualized Systems

  Resource Management (Video)
  It has been said that virtualization is time sharing done right.Virtual machines tend to hoard resources. For example, they ask for large amounts of memory but only use a small fraction of it at any time. Better use of resources often results from over provisioning. There are, however, many different resources and many different demands from VMs. The resource manager often can indirectly infer the real needs and can sometimes indirectly persuade VMs to give up control of unneeded resources.
• Computing in the Cloud — Who, What, Where, and When (Orran Krieger & Tichomir Tenev)
  

    Cloud Computing (PPT)

    Cloud Computing (PDF)

  Computing in the Cloud (Video)
  Although “cloud computing” is a common marketing term, it is possible to give it a more concrete meaning. This presentation focuses on an infrastructure for “utility computing” on top of which, all sorts of clouds can be assembled. It is argued that computing should be treated as a utility. It should not matter where one runs their application just like it does not really matter when one’s electricity is generated.The lectures cover motivation, needs, uses, and what technology is needed for them. In particular, VMware’s vCloud API will be presented. Other features, challenges, and future directions will be covered.
• One Future of Virtualization: From Phones to Smartcomputers (Prashanth Bungale & Larry Rudolph)
  

    ARM Virtualization (PPT)

    ARM Virtualization (PDF)

    Future of Virtualization (PPT)

    Future of Virtualization (PDF)

  One Future of Virtualization (Video)
  To date, most virtual machine technology has been focused on the personal computer and server markets, most of which are based on the x86 processor architecture. Most mobile phones, on the other hand, are based on the ARM processor architecture. This lecture examines the ARM ISA in detail to identify how it differs from x86 in regards to virtualization.Mobile phones are resource challenged and supercomputers are latency challenged. In each case, there are pros and cons to virtualizing these machines. In addition, there are many other (non-computing) devices that may benefit from virtualization. This final set of lectures highlights some future use cases.