Document Sample
jain Powered By Docstoc
					                                       Rethinking Pervasive Computing
                                                        Ravi Jain

Abstract. We argue that pervasive computing offers not only tremendous opportunities and exciting research
challenges but also possible negative environmental impacts. We argue for rethinking its methodology to
minimize these impacts; in particular, that design for minimizing total-lifecycle resource consumption and
residue should become a mainstream concern. Finally, in keeping with the workshop theme, we briefly mention
some computer science implications and the development of techniques to help manage resource consumption
and environmental impacts.

                                                                for landfills, incinerators or hazardous waste exports
                                                                [NSC99]. Consider that computer equipment is a
1     Introduction                                              complicated assembly of over a thousand materials, of
                                                                which many, such as lead, cadmium and mercury, are
Pervasive or ubiquitous computing has been driven by
                                                                known to be highly toxic; what may be worse, the
ambitious, exciting and noble goals – to put
                                                                effects of some of these are still unknown. Thus the
computing machines in the service of humanity,                  environmental impacts of computers have been a
instead of the other way round; to make computing as            subject of growing concern over the past decade
useful and unobtrusive as utilities like electricity and
water; to produce “calm” rather than distraction; and
to bring the benefits of computers to everyone by
                                                                Already, the growth of waste electrical and electronic
developing not only powerful, costly machines but               equipment is about 3 times that of other municipal
“tiny inexpensive ones” [Weiser93, Weiser96]. While             waste [AEA97]. The average lifespan of a computer
this vision has not been realized, much progress has
                                                                tower has shrunk from 4-6 years in 1997, and is
been made, and researchers and engineers around the
                                                                estimated to fall to 2 years by 2005. Pervasive
world are chipping away at the obstacles, slowly but
                                                                computing will only add to this “mountain of obsolete
                                                                PCs” [Watson99], both by increasing the nature and
                                                                quantity of physical devices and the rate at which they
However, little or no thought has been given to the             become obsolete.        Not only computing, but
physical final end result of pervasive computing:
                                                                communications devices and integrated computing
devices of varying size, weight and complexity, that
                                                                and communications devices are expected to
are useless, obsolete, malfunctioning, or simply
                                                                proliferate. Analysts predict that worldwide by 2005
broken – in other words, that are garbage. Further,
                                                                there will be over a billion new Bluetooth devices
these devices, by their very design and function, are           alone [Cahners00]. And the speed of innovation in
ubiquitous, massively distributed, and embedded in              wireless and PDA technology, coupled with its
numerous everyday objects and the environment.
                                                                relatively low cost, indicates decreased lifetimes for
From digital jewelry and clothing, to networked
                                                                devices. One year’s device may be replaced because
appliances, sensor networks, smart floors and
                                                                it is not WAP-enabled, and the following year the
cyberhomes, pervasive computing offers us not only a
                                                                replacement may be replaced because it is not
glittering future of convenience, comfort and                   Bluetooth-enabled.
connectivity, but possibly also a legacy of deadening
clutter and dangerous trash: plastics that do not
                                                                Generally most pervasive computing devices –
biodegrade, heavy metals that are carcinogenic, gases
                                                                although by no means all -- will have one significant
from production and incineration that are toxic, and
                                                                environmental advantage over traditional computers:
landfills that threaten generations to come.                    that they are physically smaller and inherently
                                                                consume less material. However, they have other
To gauge the extent of this possibility, it is worthwhile
                                                                disadvantages: they will be far more numerous; low
glancing at the physical waste produced by the current
                                                                cost will encourage rapid replacement; less mature
– i.e., non-pervasive – use of computers. It has been
                                                                technology will become obsolete faster; disposable
estimated that over three-quarters of all computers
                                                                versions of some devices, like disposable cell phones
ever bought in the U.S. are stored in people’s attics,          [HopOn01, Telespree01], will soon emerge; they will
basements, office closets and pantries [MCC96]. By              tend to use batteries, which often contain heavy metals
the year 2004, experts estimate there will be over 315
                                                                and are an environmental hazard in themselves, rather
million obsolete computers in the US, many destined
                                                                than fixed AC power. In addition, their small size,
Position paper, NSF Workshop On Context-Aware Mobile
And Pervasive Data Management, January 24-25, 2002.
 TelcordiaTechnologies Inc., 2002
weight, embedding in other materials and overall               be an integral part of the design process, not an
design for ubiquity will disperse them widely, making          afterthought or enhancement. This view has been
them more likely to be lost, forgotten, or simply              articulated for manufactured products in general, e.g.:
abandoned, and making proper collection, recycling or          “Very few objects of modern consumption were
disposal harder.       Finally, pervasive computing            designed with recycling in mind. If the process is
devices, to meet their goal, will be truly global,             truly to save money and materials, products must be
bringing computer environmental impacts to regions             designed from the very beginning to be recycled …”
of the world where little or none exist at present.            [McB98]. In fact, unlike current computing, where
                                                               environmental concerns were only raised after the
So far we have focused on the physical waste aspects           proliferation of computers, pervasive computing offers
of pervasive computing.           There are other              us a unique opportunity to apply the environmental
environmental impacts to consider, such as energy              consciousness raised by current technology while we
consumption. It has been estimated that computing,             are still at the start of the next wave of technology
telephony and networking equipment now accounts                proliferation. This belief motivates the title of the
for a significant fraction of the total energy                 paper: we believe it is important to rethink pervasive
consumption in the U.S. [EStar]. While mobile and              computing in the sense that minimizing total lifecyle
pervasive computing devices (such as laptops)                  environmental impact should become one of the
individually are more energy-efficient than before, the        important factors in pervasive computing design.
overall energy consumption due to such devices
continues to increase because their total number               Of course, much of pervasive computing research
increases rapidly, they integrate more sophisticated           focuses on design to minimize resource consumption,
and energy-consuming peripherals (larger displays,             mainly because of limitations of device size, weight
built-in wireless interfaces, CD-R/W etc.), and the            and capabilities that make the research interesting in
applications and system software (and even screen              the first place. However, the research is generally
savers) become ever more complex [Paradiso00].                 driven by the goal of squeezing more functionality out
Once again, in the case of pervasive computing, this           of the resources available, where the resources in turn
environmental impact is greater than current                   are limited primarily by cost, or sometimes by
computing because of the use of batteries.                     availability. At present environmental impacts that
                                                               are external to the producer, i.e., after a product leaves
We briefly mention another issue with pervasive                the factory, are not generally reflected in costs for
computing that, in our opinion, has been insufficiently        electronic devices. This is especially true in the US,
addressed: the possible proliferation of clutter. It is        although recently in Europe “extended producer
possible that just as the “paperless office” innovations       responsibility” policies are shifting the burden of
of the 1980s led in fact to far greater usage and              waste electrical and electronic equipment to the
disposal of paper [Brooke01], the ideas of calm                manufacturer [EC00]. Until this occurs, we argue that
computing inherent in the pervasive computing vision           pervasive computing design should explicitly consider
will lead to more computerized distraction and                 and minimize environmental impacts as a separate
annoyances. We omit further discussion in this paper           parameter from cost. This is not as radical as it may
due to lack of space.                                          sound; many corporations have voluntarily embraced
                                                               principles of minimizing environmental impact.
2    Design implications
                                                               We also argue that research should consider
The intent of this paper is not to present an alarmist
                                                               minimizing not only production and operation costs
view of global environmental collapse due to                   but total lifecycle impacts, i.e., choosing techniques to
computers, but to argue that computers in general, and         reduce the costs of reuse, recycling, disposal and
pervasive computers in particular, pose an
                                                               incineration. This complicates the design process, and
environmental risk that needs to, and can, be
                                                               also makes it more challenging – trading off not only
                                                               between functionality and operational or production
                                                               cost, but also total lifecycle impacts. At present there
As researchers we tend to focus on innovation, and in          are few metrics and techniques to carry out or evaluate
our view there are interesting challenges as well as           such designs, and developing these techniques for
opportunities for reducing the environmental impacts
                                                               pervasive computing is a challenging endeavor.
of pervasive computing by innovation. However, we
believe strongly that effective reduction of resource
                                                               Example Scenario. To make these ideas a little more
consumption, reuse of resources, and recycling of
                                                               concrete we sketch a scenario that is quite
materials – the reduce/reuse/recycle mantra -- needs to        conservative in the sense that it is within the realm of

technical realization. Alice leases or purchases a             Formal models of energy consumption are required
Rethink brand cell phone whose casing is made from             that explicitly consider energy instead of (or in
biodegradable or recycled plastic. It contains a large         addition to) CPU cycles in order to motivate design of
proportion of recycled electronic components and               algorithms (e.g. for data indexing or for wireless
PCBs, and connections made with lead-free solder.              computing). This is analogous to the formal models
The Rethink phone comes with a Shoe Battery                    developed to support the design and comparison of
installed and a spare that is rechargeable, when Alice         I/O-efficient algorithms [Shriver96]. Some work
walks, by means of a piezo-electric charging apparatus         along these lines is in [Ellis01].
in her shoe. The phone is sufficiently integrated,
programmable and convenient that it eliminates                 Reducing software sprawl. Operating systems and
several other devices, such as a PDA, wallet, and              system software notoriously grows in size and
various household remotes (TV, VCR, stereo, fan,               complexity with each new release, even though most
garage, car, etc.), although it may not be the only            users may not actually use many of the new features.
device Alice owns. The phone has a software radio to           This not only increases storage and energy usage but
accommodate air interface changes when Alice travels           after only a few releases can needlessly make
or when new technology is developed. It also has an            hardware obsolete. Mechanisms to discover system
open, standard API so that new applications (e.g.              capabilities, autoconfigure systems, and allow secure,
currency converters, foreign language phrasebooks)             just-in-time plug-in and assembly of required system
can be downloaded to it on demand, extending its               software components (similar to what is done in Web
function as well as life. The hardware is designed for         browsers) are required. Related to this are:
modularity and replacability, rather than being                 Dynamic application usage can extend the useful
integrated to minimize initial cost. Thus when Alice                life of a device not only by avoiding software
does decide to buy an upgrade or replacement, the                   sprawl but also allowing convenient upgrade as
phone can be easily disassembled by the store clerk or              well as application leasing.       This involves
Alice herself, and many of the physical materials                   dynamic application discovery, download,
(casing, display, keys, battery circuits, etc) are not              charging and billing, all of which has to be done
changed. The manufacturer has an incentive to avoid                 for a device with (possibly severe) resource
needless obsolescence so as to avoid disposal costs.                limitations.
Similarly, the software is in component form and is             Languages and APIs for programmability,
replaced only if necessary, and with minimum delay                  modularity and extensibility are required so that
and inconvenience. Finally, after many years, when                  system as well as application software can be
the product wears out or is actually obsolete, Alice has            designed for reuse, replacement, and upgrade.
an incentive (e.g. deposit refund or trade-in value) to
return the phone to the manufacturer or third party for        Reducing data sprawl.           The “memex” device
recycling.                                                     imagined in Vannevar Bush’s famously prescient
                                                               article [Bush45] is essentially almost a reality. The
3    Computer science issues                                   memex was designed to help a scientist manage the
                                                               vast amounts of technical information produced
Rethinking pervasive computing to minimize
                                                               worldwide by allowing creation of personalized copies
environmental impact has implications at all aspects
                                                               and hyperlinks through the research corpus. However
of system design, including computer science and data
                                                               the memex did not consider the consequences of its
engineering. Unlike current wireless cellular devices,
                                                               own success. Not only scientists and corporations but
data services and applications are expected to be a key
                                                               individuals – either personally or professionally, or
part of the value proposition for pervasive computing,
                                                               both -- now maintain relatively large quantities of
so data engineering for minimal impact is likely to be
                                                               data. Data sprawl from information unnecessarily
                                                               stored far beyond its useful lifetime consumes storage
                                                               as well as energy for search and management, and also
The scenario in the previous section touches on some
                                                               contributes to “virtual clutter” and usability and
of the issues that are worth considering, described
                                                               maintenance issues. Individuals find it easier and
below. While some of these have been investigated
                                                               cheaper to expand system resources (larger and faster
for current computing systems or the Internet, they
                                                               disks and processors or tertiary storage systems) than
raise additional challenges in a pervasive computing
                                                               to manually manage even personal information such
environment. Some techniques are oriented towards
                                                               as mail and weblogs. The proliferation of multimedia
reducing resource consumption while others are
                                                               content and the widespread use of digital personal
oriented towards increasing useful product lifetimes.
                                                               libraries (e.g. already many babies in the U.S. start out
                                                               with a web page) will make this a non-trivial issue.

We suggest that systems be designed to minimize data              [Brooke01] J. Brooke, The Paperless Office? Not by a Long
sprawl through better indexing, retrieval and                               Shot, The New York Times, April 21, 2001. See
knowledge management techniques. For instance,                    
some word processors store information in                                   1PAPE.html.
                                                                  [Bush45] V. Bush, As we may think, Atlantic Monthly. June
mysteriously inefficient ways and most store every                          1945.
document revision instead of storing just the revisions.          [Cahners00] Cahners In-Stat Group, Bluetooth Wireless
Few offer comprehensive and convenient journaling                           Chips Catapult to 1.4 Billion Units by 2005, Press
features so that some forms of data can, under user                         Release,      July      26,       2000.          See
policy controls, be self-managing (so the most recent             
revision can be located easily) or self-destroying (e.g.                    m
after a certain date specified by user policies, or after a       [EC00] European Commission, Commission tackles
certain number of copies have been made.) Email                             growing problem of electrical and electronic
attachments are needlessly copied to multiple                               waste, DN: IP/00/602, Press Release, June 13,
                                                                            2000.                                            See
mailboxes (and sometimes to multiple devices for a                
given recipient) rather than being automatically                            ction.gettxt=gt&doc=IP/00/602|0|RAPID&lg=EN
stripped and stored in a logically central location. The          [Ellis01] C. S. Ellis, The Milly Watt Project, 2001. See
list of such inefficiencies can easily be made much               
longer, mainly because the “storage is cheap” refrain             [DoE] U.S. Dept of Energy, The Energy Star Program, See
hides the human and environmental costs of managing               
obsolete or useless data.                                                   tml.
                                                                  [Estar] U. S. Environment Protection Agency, Press
Hardware/software codesign motivated not only by                            Releases: “EPA Administrator Looks to
                                                                            Telecommunications Industry for Increased
speed and cost but also by minimal environmental                            Energy Efficiency Opportunities”, 11/16/2001;
impact.                                                                     “Energy Star to Launch Label for Telephony
                                                                            Products at Consumer Electronics Show in Las
Tools to evaluate design choices in terms of total                          Vegas, January 8-11, 2002”
lifecycle costs are required.                                     [HopOn01] Hop-On Communications, See http:// www.hop-
Context-prediction techniques have some promise for               [McB98] W. McDonough and M. Braungart, The NEXT
helping manage resource consumption. Passive or                             Industrial Revolution, Atlantic Monthly, October
active prediction has been widely used in computing                         1998.
                                                                  [MCC96] Electronics Industry Roadmap, Microelectronics
to reduce CPU time e.g. in branch prediction,                               and Computer Technology Corporation, Austin,
prefetching and caching. It is recognized that user                         TX, 1996.
location prediction techniques can help minimize                  [NSC99] Electronic product recovery and recycling baseline
delay and improve QoS in wireless communications,                           report, National Safety Council, Washington, DC,
e.g. by appropriate resource reservations in                                1999.
surrounding cells. However, techniques that predict               [Paradiso00] J. Paradiso, Renewable energy sources for the
not only location but total context (presence,                              future of mobile and embedded computing, MIT
availability, running applications and state, existing                      Media Lab, Mar. 2000.
communications sessions and state, etc.) could also be            [Shriver96] E. Shriver and M. Nodine, An introduction to
                                                                            parallel I/O models and algorithms, in R. Jain, J.
used to reduce energy and bandwidth resource                                Werth and J. C. Browne (eds.), Input/Output in
consumption. We are currently engaged in developing                         Parallel and Distributed Computer Systems,
context prediction techniques for pervasive                                 Kluwer, 1996.
computing.                                                        [Telespree01] Telespree, Inc,
                                                                  [Watson99] T. Watson, “USA sitting on mountain of
                                                                            obsolete PCs", USA Today, June, 22, 1999.
Acknowledgments. Many thanks to Brian Coan and                    [Weiser93] M. Weiser, Some Computer Science Problems
John Wullert of Telcordia for comments on a draft of                        in Ubiquitous Computing, Communications of the
this article.                                                               ACM, July 1993.
                                                                  [Wesier96] M. Weiser and J. S. Brown, The coming age of
                                                                            calm technology, Xerox PARC, Oct. 1996. See
References                                                                  endnote.htm

[AEA97] AEA Technology, Recovery of WEEE, Economic
        and Environmental Impacts, June 1997.


Shared By: