
Internet Suspend/Resume
Michael Kozuch, Intel Research Pittsburgh
M. Satyanarayanan, Carnegie Mellon University and Intel Research Pittsburgh
Abstract
We identify a new capability for mobile computing that
mimics the opening and closing of a laptop, but avoids phys-
ical transport of hardware. Through rapid and easy person-
alization and depersonalization of anonymous hardware, a
user is able to suspend work at one machine and to resume
it at another. Our key insight is that this capability can be
achieved by layering virtual machine technology on a dis-
tributed file system. We report on an initial implementation
and describe our plans for improving efficiency, portability,
and security.
1 Introduction
When you close a laptop, its state is suspended. You can
travel a great distance and open the cover many hours or
days later; the laptop’s execution state is restored precisely
as it was at suspension. Can we achieve this capability with-
out physically carrying the laptop? In other words, can I
logically suspend a machine at one Internet site, then travel
to some other site and resume my work there on another
machine? Can state restoration be as swift as it is when you
open a laptop? We call this hypothetical capability Internet
Suspend/Resume (ISR).
If feasible, ISR would support mobile computing with-
out hardware being physically carried. For example, one
could pick up a laptop at a car rental agency much as one
rents a cell phone today. Or, one could imagine a laptop be-
ing built into every seat in an aircraft for use during flights,
much as video players are today. The essential capability
needed for these scenarios is the ability to effortlessly save
and restore computing state — in other words, to rapidly
and easily personalize and depersonalize anonymous hard-
ware. Of course, the term “laptop” is used here only for
illustration. This work applies equally well to any other
form of computing hardware, from desktops to handheld or
wearable computers.
The key insight offered by this paper is that ISR can be
realized by combining two off-the-shelf technologies: vir-
tual machine (VM) technology and distributed file systems.
We report on a simple proof-of-concept implementation in
this paper. We also identify many important improvements,
and describe our plans for building a more robust, efficient,
and complete prototype.
2 Design Overview
ISR is foreshadowed by location transparent distributed
file systems such as AFS [10] and Coda [16]. If a user re-
stricts all his file accesses to such a file system (including
placing his home directory in it), he will see identical file
state at all clients. He can log in to any client, work for a
while, log out, move to any other client, log in, and con-
tinue his work. There are at least two ways in which this
capability falls short of ISR. First, only persistent state is
saved and restored. Volatile state, such as the the execu-
tion states of interactive applications, is not preserved. The
second shortcoming follows from the first. From a user’s
viewpoint, the “suspend” and “resume” steps are consider-
ably more complex, heavyweight and slow than closing and
opening a laptop.
A virtual machine monitor (VMM) [6] cleanly encapsu-
lates all volatile execution state of a VM. The operating sys-
tem that executes within a VM is referred to as a guest oper-
ating system. A VMM typically maps the volatile state of its
VMs to files in the local file system of its host. When a VM
is suspended, the corresponding files are updated to reflect
volatile state at the point of suspension. If these files are
copied to a remote host with similar hardware architecture,
a VMM on that host can resume the VM. In other words, the
VM, including the volatile state of the guest OS, has been
migrated.
In contrast to the well-known difficulties of process mi-
gration [4], VM migration is simpler because volatile exe-
cution state is better encapsulated. Because migration takes
place at the machine level, some problems that have his-
torically presented challenges for process migration are not
present, such as managing open file descriptors cached by
applications. Another historic challenge, migrating network
connections, is manageable because modern desktop oper-
ating systems provide features such as ACPI [1] compliance
to accomodate laptop mobility. To the guest OS, a VM mi-
Proceedings of the Fourth IEEE Workshop on Mobile Computing Systems and Applications (WMCSA’02)
0-7695-1647-5/02 $17.00 © 2002 IEEE