Use Phase Signals to Promote Lifetime Extension for Windows PCs

Feb 19, 2009 - LEARN ABOUT THESE METRICS ... form of intuitive signals indicative of both hardware and software quality can in conjunction with consum...
0 downloads 0 Views 129KB Size
Download PDF
Environ. Sci. Technol. 2009, 43, 2544–2549

Use Phase Signals to Promote Lifetime Extension for Windows PCs STEWART HICKEY,* COLIN FITZPATRICK, MAURICE O’CONNELL, AND MICHAEL JOHNSON Department of Electronic and Computer Engineering, University of Limerick, Limerick, Ireland

Received July 24, 2008. Revised manuscript received February 2, 2009. Accepted February 3, 2009.

This paper proposes a signaling methodology for personal computers. Signaling may be viewed as an ecodesign strategy that can positively influence the consumer to consumer (C2C) market process. A number of parameters are identified that can provide the basis for signal implementation. These include operating time, operating temperature, operating voltage, power cycle counts, hard disk drive (HDD) self-monitoring, and reporting technology (SMART) attributes and operating system (OS) event information. All these parameters are currently attainable or derivable via embedded technologies in modern desktopsystems.Acasestudydetailingatechnicalimplementation of how the development of signals can be achieved in personal computers that incorporate Microsoft Windows operating systems is presented. Collation of lifetime temperature data from a system processor is demonstrated as a possible means of characterizing a usage profile for a desktop system. In addition, event log data is utilized for devising signals indicative of OS quality. The provision of lifetime usage data in the form of intuitive signals indicative of both hardware and software quality can in conjunction with consumer education facilitate an optimal remarketing strategy for used systems. This implementation requires no additional hardware.

1. Introduction Personal computer (PC) lifetime extension through reselling in C2C secondary markets has been identified as an advantageous end of life (EOL) strategy from an environmental perspective (1-3). Matching the capabilities of used systems to the needs of second-hand buyers can substitute for the sale of new systems and consequently reduce manufacturing overhead in this product domain. There are also favorable economic and social implications associated with the practice of reselling these systems in the C2C marketplace. However, consumer perception of value of used PCs must increase so that more favorable conditions exist for secondary market activity as consumers are responsible for the flow of used computers into the market and they also represent the potential buyers of used systems (3, 4). From the demand side, consumer willingness to pay (WTP) is the maximum amount the consumer is prepared to give for the purchase of the used system (4). Falling prices of new computers is obviously an opposing force to the purchase of second-hand machines in this regard, and PC prices continue to drop as technology improves and rivalries among the major manufacturers intensifies. This is an inherent * Corresponding author phone +353(0)879493971; +353(0)61-338176; e-mail: [email protected]. 2544

9

fax:

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 7, 2009

characteristic of the product group. However there are likely to be other opposing market forces. One obvious barrier to trade is the issue of seller credibility (3, 5). For a transaction to take place, the buyer must have sufficient trust in the information provided by the seller. This applies to transactions concerning all secondary market durable goods (5). Other opposing forces may be the belief that used systems are inferior in quality to new equivalents and used systems may possibly not coincide with consumer preferences (3). In order to have the optimal environmental outcome, these preferences and beliefs must be understood and a countermeasure put in place to ensure most favorable conditions exist for secondary market activity. From the supply side, consumer willingness to receive (WTR) is the minimum amount the consumer is ready to accept as a compensation for the sale of the used system (4). Technology features, aesthetic appearance, performance per unit economic value, level of environmental impact, and emotional value have been identified as factors for secondary durable products (6). Understanding how consumers’ value used PC products is necessary in order to incorporate an appropriate counter-measure. Time-sensitive consumer electronics products such as PCs have been reported to lose value at rates in excess of 1% per week, and the rate increases as these products near the end of their life (7). Therefore lack of timely involvement of consumers on the supply side can be equally as devastating for secondary markets and will inevitably result in product obsolescence. This is a significant concern considering the number of PCs globally is expected to increase from 592 million in 2002 to more than 4 billion in 2020 (8). Gartner has found that demand for used computers continues to exceed supply in developing regions such as Eastern Europe, the Middle East, Africa, Latin America and parts of the Asia-Pacific region (9). Consumer WTP does not seem to be the issue in these regions and it is consumer WTR that is the likely factor preventing the flow of quality used PCs to these potential lucrative markets. In developed economies both consumer WTR and consumer WTP together have been shown to stifle efficiency in these markets (3). Signaling is a strategy that can serve to positively influence both consumers WTP and WTR. It is a means whereby lifetime usage data can be utilized to facilitate greater information exchange between buyers and sellers thus serve to increase market efficiency. The importance of signals indicative of both hardware and software reliability for use in C2C transactions has been established in previous work (3). This paper complements that work by demonstrating how signaling of both hardware and software quality may be achieved with a PC incorporating a Microsoft Windows OS. This can in conjunction with consumer education facilitate an optimal remarketing strategy for used Windows PCs.

2. Review of Literature The environmental impacts incurred in PC production are significantly exacerbated by their short life spans which have been recognized as very short among durable goods (10-14). Lifetime extension has been shown to be an environmentally and economically superior alternative to recycling when it comes to the life cycle management of these systems. A range of strategies which facilitate lifetime extension of PCs have been recommended with reselling perceived as the best EOL option from an environmental perspective as it delays and often substitutes the requirement for a new system (3, 15). It has recently been proposed that life cycle data qualifying the usage of durable consumer products in their original 10.1021/es8020638 CCC: $40.75

 2009 American Chemical Society

Published on Web 02/19/2009

lifecycle environment can play a significant role in remarketing of these products (3). Conventional approaches to lifetime data acquisition have utilized a life cycle data acquisition (LCDA) unit for monitoring and storage of operating and environmental conditions. This unit is an independent, self-powered data-recording device equipped with various sensors and a microcontroller to monitor the environmental loads experienced by the system at programmed intervals (16-19). In Europe, the recent Product Lifecycle Management and Information Tracking using Smart Embedded systems (PROMISE) initiative, has shown that the adoption of LCDA technologies in high end electromechanical systems can result in positive environmental and economic implications for producers (20). The use of LCDA technologies have further been proposed for use in consumer electronics durables (18, 21). However, when considering the potential integration of LCDA technologies in client PC systems, financial incentives for original equipment manufacturers (OEMs) are perceived as somewhat less obvious. This was evident in the results of the European Commission funded Environmental Life-cycle Information Management and Acquisition (ELIMA) project for consumer products in 2005, where return on investment due to costs incurred from the incorporation of the LCDA modules was notably absent despite the presence of considerable environmental and user knowledge gains (22). The use of existing onboard diagnostic sensors has since been proposed to address the economic challenge of condition monitoring in desktop PCs (3). When considering the potential role product usage information can play in remarketing, in addition to acquiring usage data, it is also necessary to determine the optimum format this information should take to facilitate ease of qualification for potential secondary buyers. Life consumption monitoring (LCM) has been proposed to assist in EOL decision making for electronic systems (23). LCM is a method of monitoring parameters critical to a system’s health and converting the measured data into life consumed. This process consists of (1) monitoring the critical parameters of a products usage environment, (2) simplifying data, and (3) modeling physics of failure analysis. Monitoring the critical parameters of the product’s life cycle environment involves the continuous monitoring of the assembly, storage, handling, and use of the product including the severity and duration of application and operational loads endured in each phase. Simplification of data consists of a conversion of the monitored data into a form compatible with the input requirements of reliability assessment models. Physics-offailure analysis involves the determination of the cumulative damage accumulation in the product due to various failure mechanisms induced by the monitored loads (23). The concept of LCM has already been successfully applied to mechanical systems. For example, LCM is used in the automotive industry for engine oil monitoring. Time, temperature, and parameters related to engine usage are processed with the appropriate degradation models to estimate the approximate remaining life of oil (24). LCM has also been applied to electronic modules utilized in aerospace applications (25). The concept has further been proposed for use in electronics durables, with a methodology already developed that focuses on the estimation of accumulated damage of solder joints in electronics (26). It is probable that a LCM computation on the use phase of a PC in the form of mean time to failure (MTTF) or equivalent metric would hold considerable value for potential buyers of secondary systems, facilitating ease of secondary PC qualification regardless of consumer level of expertise. However, LCM is neither practical or feasible for assessing an entire PC system, as there are far too many different components with different failure modes to estimate, with

any level of confidence, a measure of life consumed (Further discussion on this topic is given in the Supporting Information (SI)). Nonetheless, the provision of life cycle data in the form of intuitive signals remains an imperative measure to instil secondary buyer confidence in the value of used equipment. This research seeks to enhance secondary market activity by exploiting such available usage data and providing a means by which equivalent usage metrics can be represented. An overview of the sensor hardware incorporated in current desktop PCs is initially given in Section 3. This provides a background on existing technologies required to facilitate the signaling methodology proposed. Section 4 describes a case study detailing how signaling metrics indicative of PC hardware and OS quality can be generated in Windows PCs. All the proposed signals attempt to enhance consumer perception of PC system quality and are generated using no additional hardware.

3. Reliability Indicators for Secondary Use Markets Before the signaling methodology is presented, it is first necessary to introduce the embedded technologies in modern PCs that can facilitate access to usage and reliability parameters required for usage profile qualification. Sensor Hardware. In recent years, primary computer components have been designed with built-in sensors that monitor these components as to how they are being used. Temperature sensors are currently incorporated in all modern processors and certain high performance graphics cards. The rate of hardware faults occurring in the useful life of electronic components is tightly coupled with increasing temperature (27). Voltages of primary system components are also software visible. Overvolting of CPUs is common among power users as a means of enhancing stability while overclocking (overclocking is the term used for changing a system’s configuration to that it runs at a faster speed than that recommended by the manufacturer). By recording the voltage supplied to key components, it is possible to detect voltage spikes or deviations which would indicate degradation of components. For example, the deviation of system voltages from nominal is an important indicator of the health status of the power supply. Certain PC fans are equipped with fan management ICs which monitor fan speed. The most basic fan failure detection is implemented using discrete components. Some available fan management ICs can monitor the fan’s commutation pulses and assert an alarm signal when no pulses are detected (28). Modern hard disk drives (HDDs) currently incorporate SMART diagnostics that permit the end user to evaluate the probability of drive failure at any given instant. Predictable failures are characterized by degradation of a certain attribute over time before the disk drive fails. Mechanical failures and other certain electronic failures are considered predictable because they show a degree of degradation before failing (29). Examples of attributes set by drive manufacturers include read error rate, throughput performance, power-on hours, uncorrectable sector count as well as a host of others (30). A recent survey carried out on a large population of disk drives has established that scan errors, reallocation counts, offline reallocation counts, and probational counts have the most significant impact on disk drive reliability (31). The diagnostic technologies mentioned above have been incorporated primarily as health monitors to provide feedback to users on the current operating health of their systems. This feedback is particularly of interest to power users where “over clocking” is common practice causing their system components to run at higher temperatures than intended by the manufacturer. Software Systems. In addition to embedded hardware instruments, operating systems (OSes) facilitate access to a VOL. 43, NO. 7, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2545

range of supplementary parameters that may be used to qualify a usage profile for an individual system (Windows Management Instrumentation (WMI) is a technology in current Windows distributions that facilitates access to voltage, processor usage, uptime, and a host of other raw data operating attributes through advanced configuration and power interface (ACPI)). These are described in the subsequent paragraphs. Operating Time. The total system operating time may vary considerably for systems of the same age and therefore is an important attribute when signaling quality. All the reliability prediction methodologies predict that operating hardware failure rate is greater than that of non operating (32). Primary system component manufacturers also recognize that reliability depends on age of their products. Modern HDDs are specified with a MTTF of at least 500 000 h for the first 5 years of operation corresponding to 1.7 failures per year. Beyond 5 years, this failure rate rapidly increases (33). The lifetime of a typical fan is specified at 50 000 h corresponding to 5.7 years (34). A consequence of fan failure is overheating which has a detrimental affect on PC system components (35). Leaving a PC powered on also increases the probability of damage due to power surges and spikes. Electrolytic capacitors, which are fundamental components in power supplies, also tend to dry up with time affecting the operation of the power supply (36). A fault in the power supply can cause erratic voltages to propagate through the system and affect the other primary system components. System operating time has also been shown to negatively affect reliability of OSes (37). Power On/Off Cycles. Large temperature cycles occurring as a consequence of powering up and down, or going into low power or stand-by mode, can impact both solid state and electromechanical system component reliability. Cycle failure estimates for the HDD and power supply indicate the allowable cycles for these components at 75 000 and 20 000 cycles, respectively (38). Crash Dump/Event Log Information. Thus far, parameters that can be utilized to develop signals indicative of hardware reliability have been described. Parameters indicative of software reliability must also be identified for signaling. This is reflected in a previous customer survey with 81% considering the “number of system crashes in the previous lifetime” as a very important criterion when considering the purchase of a second-hand PC (3). Current Windows distributions record specific OS failure data in a “mini-dump” file with each crash or blue screen generated by the system. These mini-dumps are a partial snapshot of the computer’s state at the time of crash and in their simplest form contain a list of loaded drivers, the names of the binaries that were loaded in the computer’s memory at the time of crash, the processor context for the stopped process, and process information and kernel context for the stopped process and thread and also a brief stack trace (39). It is probable that the most effective solution to signaling software quality is to give second-hand buyers crash information in the format they desire, i.e., provision of minidump information in quantitative form. However, minidumps will not be recorded for every crash. In an instance where the kernel code or data structures required to write the crash dump have been corrupted, no crash mini-dump will be generated (40). Certain hardware crashes, for example a thermal emergency in the CPU or failing power supply, will also cause the computer to restart immediately without creating a memory dump file. Therefore, it may be misleading to provide signals communicating quantitative crash data to consumers. This is where event log information may serve useful. In addition to OS crash logging, Windows systems provide analogous logging capability for hardware, software, and 2546

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 7, 2009

configuration error events. Logged events include information regarding the type of event and also the date and time of occurrence (41). In Windows XP, events are categorized as being of type information, warning, or error with the OS placing these events into three separate logs. (1) The application log, (2) The security log, and (3) the system log. The application log contains events logged by programs. For example, application crashes and application hangs would both be recorded in the application log. The system event log records events for Windows OS components. For example, a driver failing to load during start-up would be recorded in the system log. Failure modes from event logs are used for software reliability and dependability analysis (41) but provision to consumers in a secondary market could also be used to enhance their perception of system quality. When considering the potential value of event log information in a secondary market transaction it is clear that they are presently neither readily accessible nor in a format that potential secondary consumers can comprehend. Corporate and academic research communities attach high importance to complete error and erroneous behavior retrieval but it is reasonable to assume that the average consumer will not. Intuitive signals are necessary to raise secondary buyer confidence in relation to the quality of installed software. Signaling Methodology. The following sections detail the steps of the signaling methodology for PC systems (A diagram of the conceptual flow is given in the SI). This proposal combines internal sensing with use profiles and onboard information processing to yield useful indicators of reliability for secondary buyers. The steps include (1) monitoring usage loads/indicators representative of system usage/reliability degradation, (2) signal metric development, and (3) signaling metric provision. Monitoring Usage Loads/Indicators representative of System Usage/Reliability Degradation. Parameters indicative of system reliability degradation can be classified as hardware or software types. Examples of parameters indicative of PC hardware degradation include temperature and voltage overages/fluctuations and those degradation indicators attainable from HDD SMART diagnostics. Variables indicative of OS quality degradation include crash dumps and also those software errors contained in system event logs. This is not an exhaustive list for usage qualification. The loads, symbolic of “mileage”, accumulated by the system hardware could be expanded to include other variables for example humidity and vibration. These loads can also have a large influence on the lifetime of computer products (23). The future development of SMART diagnostics for PC power supplies to enable recognition of associated failure precursors’ deterioration, could qualify the use of supplementary indicators as usable parameters for signaling. Signal Metric Development. Signaling metric development is needed to collate and represent the abovementioned parameters in a meaningful way to consumers. Certain usage variables will be more easily represented then others. For example operating time will not require any processing as most consumers can already identify with its significance. On the other hand usage parameters representative of HDD and OS quality will require considerable additional processing to make them more comprehensible. Usage loads associated with hardware reliability degradation for example temperature, and voltage may be graphically presented by a series of peaks and valleys in a specific time domain. However to provide usage loads over the entire lifecycle of a PC in such a form may cause confusion among secondary buyers. A means of data reduction or screening is therefore required to condense these load histories. Data simplification methodologies condense load histories without sacrificing important damage characteristics. Existing methods for data

simplification include data reduction and cycle counting (23). These strategies are used in fatigue analysis but also could be used as a means of generating comprehensible signals that consumers can understand and derive their own interpretation of quality from. It will also be necessary to state the manufacturer of the system component (where applicable) for each signal generated. Consumers often associate assured levels of reliability with certain brands and therefore usage signals alone will not suffice for a particular system component. Signal Metric Provision. The previous sections have detailed the preliminary steps necessary to facilitate development of signals for secondary systems. Signaling metric storage and provision are beyond the scope of the supplemented case study, however they still need to be addressed. Recommendations for these issues are purely conceptual at this stage and are given in the following paragraphs. The integration of an onboard memory module in computer systems is necessary to enable reliable storage of lifetime degradation information. In the case of lifetime temperature/voltage loads, experienced by the system, a time series log record is necessary. In the instance of SMART HDD data, status updates at periodic intervals can suffice for implementation. The memory module and associated bus communication infrastructure to onboard diagnostic instruments also need to be intrusion proof to ensure signaling information is not tampered with. The mechanism by which signals can be read (by both sellers and buyers) is another important issue. Signal provision can be achieved by the inclusion of a core OS software utility in current Windows distributions. The signals generated by this utility could be uploaded through a secure web portal for online validation before provision to secondary market players. Radio frequency identification (RFID) is another technology also ideally suited to signaling information retrieval and could be particularly beneficial if signaling was adopted in the business to business (B2B) and business to consumer (B2C) market spaces. A recent report by the Green Electronics Council (GEC) concerning electronics design to enhance reuse/recycling value has termed “triage” as the “inventorying, sorting and as appropriate, the testing, of incoming material in order to route into the selected business activities” (42). The signaling methodology proposed in this paper utilizing RFID as a means of signal extraction could be instrumental in ensuring positive economic feedback is received in future triage operations for computer systems.

4. Case Study This section describes the extraction and collation of a usage data set that can potentially serve as valuable signals of quality when it comes to remarketing a used computer. This is essentially Steps 1 and 2 of the signaling methodology previously proposed. Step 3 is reserved for future work. Signals presented here focus on characterizing a usage profile for a PC based on lifetime temperature data. Development of signals pertaining to HDD and OS reliability are discussed in the SI. The system used for the study incorporated a 3 GHz Pentium 4 Prescott processor, 120GB Samsung HDD and 1024 megabytes of RAM (discussion on the system’s temperature sensor response, sampling interval, and software configuration are provided in the SI). User Classification. The following experiments attempt to translate system temperature profiles into meaningful signals from which consumers can understand and derive their own interpretation of quality. Three benchmarks were predefined to represent three different user profiles and are detailed in SI Table 1. Each benchmark constitutes a mix of typical web browsing and office applications on one extreme and a set of power user applications on the other. Figures

FIGURE 1. Light user profile.

FIGURE 2. Power user/gamer profile. 1 and 2 are experimental results and represent the temperature profiles for the light and power user profiles specified (The intermediate user profile is given in the SI). When testing the light user session, the Pentium IC’s temperature resides between 45 and 50 degrees for approximately 70% of the testing period. Operating temperature values never exceed 65 degrees when simulating this type of user. This contrasts with the power user profile where the Pentium exhibits temperature values between 60 and 65 degrees for almost 60% of the operating period. When using a simple Arhenius based rule of thumb, a margin of approximately 10 degrees between two user profiles has been said to double system reliability (43). Application induced temperature cycles are also more prominent in the case of the Power User’s profile (see SI). Secondary PC Selection. When considering how temperature profile data can be used to assist in the consumer decision making process for second-hand PCs, a means of uniform representation is necessary for individual systems. Different solid state components are likely to exhibit dissimilar operating temperature profiles as a consequence of variability in design/manufacturing processes. Considerably more variability will be introduced depending on form factors and hardware configurations. A means of uniform representation, to facilitate ease of decision making for consumers is required for implementation in real world markets. One means of achieving this is the qualification or provision of temperature data in the form of operating temperature bands. A temperature band may be defined as the temperature interval within which a certain intensity of use will occur. These operating bands could be provided by OEMs for system CPU and GPU components and used to qualify usage intensity for individual systems. For example, those temperature intervals indicated by 40-50 degrees and 60-70 degrees could qualify as the lower- and upper-most bands, respectively, for our system. A signal comprised primarily of temperature data in band 1 could be indicative of light or office use for that particular system. Alternatively, a signal VOL. 43, NO. 7, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2547

largely composed of temperature data within the uppermost band would be indicative of power use. A signal composed primarily of light usage bands would obviously be the most enticing to potential buyers. Signals for the temperature profiles attained previously are given in the SI.

5. Discussion The previous section demonstrated the development of secondary market signals for computer systems. These final sections provides further interpretation of the results and clarifies the status of what needs to be done with current sensors and software to realize the proposed signaling methodology. Collation of lifetime temperature data from a system processor was demonstrated as possible means of qualifying hardware quality for an individual system. Computer system manufacturers obviously need to further investigate temperature sensor response characteristics for this purpose. On the basis that sensor suitability is established, an important hardware limitation needs to be addressed. Although all modern processors contain temperature sensing devices, the issue of temperature access remains a barrier to wide scale implementation. Only a minority of motherboards contain the diagnostic module which enables access to CPU temperature readings (3). There are two ways of addressing this issue. One solution could be a standard whereby all motherboards would be equipped with a diagnostic module which enables access to CPU temperature readings. However, a superior solution would be a standard specifying the inclusion of a digital thermal sensor (DTS) sensor, like those incorporated in current dual and quad core packages, in all new processors (44). In this case temperature could be accessed directly from the CPU thus require no additional hardware (This is only the case with modern processors that utilize an analog thermal diode for temperature acquisition; see the SI). It is clear that the signals derived for the HDD and OS (SI) require further filtering and simplification in order to make them usable in secondary markets. In relation to the SMART HDD diagnostics, an industry wide standard for SMART is required first to initially ensure correct implementation on computer platforms and subsequently to facilitate development of more intuitive signals for non technical users. The information available from new generation of SMART technology recently incorporated in solid state drives (SSD’s) can potentially address the complexity issue associated with current SMART indicator status representation. SSD’s have no moving parts therefore many of the parameters monitored by existing SMART technology are not applicable. Indeed, it is speculated that write/erase cycles are the only real failure mechanism present in solid-state storage (45). New generation SMART technology has the capability to detect these attributes and report system usage to the host system (45). In addition to providing more perceptive feedback to users concerning drive usage and failure probability, the transition to widespread incorporation of SMART enabled SSD technology in future desktops can permit the development of more intuitive signals for buyers of used equipment. Signals indicative of software reliability can be made more intuitive by adopting the approach taken in Windows Vista. Vista has incorporated “Reliability Monitor” analysis tool (SI), that provides a system stability overview and details events that impact reliability. It calculates a “stability index” shown in a system stability chart over the lifetime of the system (46). Although the stability index provided by Vista’s reliability monitor primarily serves as feedback for maintenance purposes, it epitomizes a metric that consumers can understand, i.e., the deviation from an absolute value reflective of system stability. It could also be used to instill confidence 2548

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 7, 2009

in the quality of installed software and convince buyers that second-hand equipment can be value for money.

6. Future Work It will ultimately be necessary to ascertain the value that signals can hold in the eyes of consumers. In the C2C market case it must be established whether signals based around the ones devised in the case study can serve to reassure consumers in the quality of second-hand equipment. Signals need not necessarily coincide with consumer preferences. In the case where the relevance of a signal is not apparent, educational cues can be used to make the distinction more clear. In the automobile market consumers do not need to know the internal mechanical workings of a diesel engine to acknowledge the vehicles worth for roughly two hundred thousand miles. In the same fashion, the provision of life cycle data in the form of intuitive signals, qualifying the usage of a PC in its original life cycle environment can serve to raise consumer confidence in the value of the used equipment. Standardization of signals across the computer industry can be achieved by a standardization body. Their adoption can be encouraged by their incorporation in eco-labels or inclusion in the relevant EuP implementing measure.

Acknowledgments This work was supported by the Irish Research Council for Science Engineering and Technology (IRCSET) and in part by Microsoft Research through the European PhD Scholarship Programme. We thank the manuscript reviewers for their valuable comments on this paper.

Supporting Information Available A discussion about the usability of life consumption monitoring for personal computers and sections providing results, the case study, and hardware and software technologies for signaling. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Williams, E.; Kahhat, R.; Allenby, B.; Kavazanjian, E.; Kim, J.; Xu, M. Environmental, social, and economic implications of global reuse and recycling of personal computers; Environ. Sci. Technol. 2008. (2) Kahhat, R.; Kima, J.; Xua, M.; Allenby, B.; Williams, E.; Zhanga, P. Exploring e-waste management systems in the United States. Resour., Conserv. Recycling 2008, 52 (7), 955–964. (3) Hickey, S.; Fitzpatrick, C. Combating adverse selection in secondary PC markets. Environ. Sci. Technol. 2008, 42 (8), 3047– 3052. (4) Michaud, C.; Lierena, D. An economic perspective on Remanufactured Products: Industrial and Consumption Challenges for Life Cycle Engineering. 13th CIRP International Conference on Life Cycle Engineering, Leuven, Belgium. (5) Akerlof, G. A. The market for ‘lemons’: Quality uncertainty and the market mechanism. Q. J. Econ. 1970, 84 (3), 488–500. (6) Van Nes, N.; Van Nes, N.; Cramer, J.; Stevels, A. A Practical Approach to the Ecological Lifetime Optimization of Electronic Products Proceedings. EcoDesign 99: First International Symposium On Environmentally Conscious Design and Inverse Manufacturing., 1999108-111. (7) Sundin, E. “Product and Process Design for Successful Remanufacturing; Linko¨pings Universitet, SE-581 83: Linko¨ping, Sweden, 2004. (8) SMART 2020: Enabling the Low Carbon Economy in the Information Age; The Climate Group: London, 2008; p 19. (9) Escherich, M. Thriving Secondary PC Market Puts Old PCs to Good Use, Gartner, 2005, 23 Aug. http://www.gartner.com/ resources/130000/130064/thriving_second.pdf (Accessed 0102-09). (10) Williams, E.; Ayres, R.; Heller, M. The 1.7 kg microchip: energy and chemical use in the production of semiconductors. Environ. Sci. Technol. 2002, 36 (24), 5504–5510.

(11) Jonbrink, A. K. EUP Lot 3 “Personal Computers and Computer Monitors Final Report (Tasks 1-8),” www.ecocomputer.org (Accessed 2007, 27 Aug). (12) Williams, E. Energy Intensity of Computer Manufacturing: hybrid assessment combining process and economic input output methods. Environ. Sci. Technol. 2004, 38 (22), 6166–6174. (13) Murphy, C. F.; Kenig, G. A.; Allen, D. T.; Laurent, J. P.; Dyer, D. E. Development of Parametric Material, Energy, and Emission Inventories for Wafer Fabrication in the Semiconductor Industry. Environ. Sci. Technol. 2003, 37 (23), 5373–5382. (14) Krishnan, N.; Boyd, S.; Somani, A.; Raoux, S.; Clark, D.; Dornfeld, D. A hybrid life cycle inventory of nano-scale semiconductor manufacturing. Environ. Sci. Technol. 2008, 42 (8), 3069–3075. (15) Williams, E.; Ruediger, K. Today’s markets for used PCS-And ways to enhance them. In Computers and the Environments Understanding and Managing their Impacts; Williams, E., Ruediger, K., Eds.; Kluwer Academic Publishers: London, 2003; Vol. 14. (16) Scheidt, L.; Shuqiang, Z. An Approach to Achieve Reusability of Electronic Modules. IEEE International Symposium on Electronics and the Environment, 1994; pp 331-336. (17) Klausner, M.; Grimm, W. M.; Hendrickson, C.; Horvath, A. Sensor-Based Data Recording of Use Conditions for Product Takeback. Proceedings of the 1998 IEEE International Symposium on Electronics and the Environment. 1998; pp 138-143. (18) Klausner, M.; Grimm, W.; Hendrickson, C. Reuse of Electric Motors in Consumer Products: Design and Analysis of an Electronic Data Log J. Ind. Ecol. 1998. (19) Simon, M.; Bee, G.; Moore, P.; Jun-Sheng, P.; Changwen, X. Life Cycle Data Acquisition Unit-Design, Implementation, Economics and Environmental Benefits. Proceedings of the 2000 IEEE International Symposium on Electronics and the Environment (ISEE), 2000; pp 284–289. (20) PROMISE. Product Lifecycle Management and Information Tracking Using Smart Embedded Systems. http://www.promise. no (Accessed 07-12-08). (21) Simon, M.; Pu, J.; Moore, P. The Whiteboxscapturing and using product life cycle data. In Life Cycle Design- 5th CIRP Seminar; CIRP: Stockholm, 1998; pp 161-170. (22) Seminar, E. Product Life Cycle Technologysresults and Implications for Business Strategy, ELIMA Seminar; ELIMA: Vienna, Austria, 2005. (23) Ramakrishnan, A.; Pecht, M. G. A life consumption monitoring methodology for electronic systems, Components and Packaging Technologies. IEEE Trans. Compon., Packag. Technol. 2003, 26 (3), 625–634. (24) Sarkar, R.; Calcut, B. D.; Linden, J. L. Real time life models for automatic transmission fluids 6920412 Patent Storm 2005, (10765239).http://www.patentstorm.us/patents/6920412/fulltext. html. (25) Mathew, S.; Das, D.; Osterman, M.; Pecht, M.; Ferebe, R. Prognostics assessment of aluminum support structure on a printed circuit board. J. Electron. Packag. 2006 >, 128 (4), 339– 345. (26) MishraS. GanesanS. PechtM. JingsongX. Life Consumption Monitoring for Electronics Prognostics. Aerospace Conference, 2004. Proceedings2004; 5, pp 3455-3467. (27) Lall, P.; Pecht, M. G.; Hakim, E. B. Influence of Temperature on Microelectronics and System Reliability; CRC Press: Boca Raton, FL, 1997.

(28) Paparrizos, G. Detecting Failure in DC Cooling Fan 2005, http://www.electronicproducts.com/ShowPage.asp?FileName) microchip.may2005.HTML (Accessed 01-02-08). (29) Vichare, N. M.; Pecht, M. G. Prognostics and health management of electronics. IEEE Trans Compon., Packag. Technol. 2006, 29 (1), 222–229. (30) Get SMART for Reliability; Seagate Technology, Inc.: Scotts Valley, CA, 1999. (31) Pinheiro, E.; Weber, W. D.; Barroso, L. A. Failure Trends in a Large Disk Drive Population 5th USENIX Conference on File and Storage Technologies (FAST), 2007. (32) Goel, A.; Graves, R. J. Electronic system reliability: collating prediction models IEEE Trans. Device Mater Reliab. 2006, 6 (2). (33) Cole, D. Energy consumption and personal computers. In Computers and the Environment-Understanding and Managing their Impacts; Ruediger, K., Williams, E., Eds.; Kluwer Academic Publishers: London, 2004; p 154. (34) Mercer, K., et al“Does Power Cycling a Desktop Computer harm it?” A review of the literature, ” New Zealand Energy Efficiency and Conservation Authority, www.energywise.co.nz/content/ ew_business/office_equipment/Power_Cycling.html, Accessed 01-02-09. (35) Linc IT. PC Shutdown-to Be or Not to Be; Commonwealth Department of Primary Industries and Energy, 1995. (36) Gasperi, M. L. Life Prediction Model for Aluminum Electrolytic Capacitors. IEEE Industry Applications Conference1996313471351. (37) Buckley, M. F. SiewiorekD. P. A Comparative Analysis of Event Tupling Schemes The Twenty-Sixth Annual International Symposium on Fault-Tolerant Computing (ftcs), 1996; p 294. (38) Gianni, R.; van der Linden, P.; Arreola, O.; Hui-Mei, C.; Goyal, S.; Marks, J.; Nowshadi, S.; WhatleyR. Thermal stress relief with power management [desktop computer systems]. Proceedings of the 1997 IEEE International Symposium on Electronics and the Environment, 1997; pp 215-217. (39) Archana, G.; Viji, G.; David, P. Windows XP Kernel Crash Analysis Proceedings of the 20th conference on Large Installation System Administration Conference, Washington, DC, 2006. (40) Russinovich, M. E.; Solomon, D. A. Microsoft Windows Internals, Fourth Edition: Microsoft Windows Server 2003,Windows XP, and Windows 2000; Microsoft Press: Redmond, WA, 2004. (41) Simache, C.; Simache, C.; Kaaniche, M.; Saidane, A. Event Log Based Dependability Analysis of Windows NT and 2K Systems Dependable Computing,2002.Proceedings. 2002 Pacific Rim International Symposium on, 2002; pp 311315. (42) Closing the Loop: Electronics Design to Enhance Reuse/Recycling Value Discussion, Draft Report; Green Electronics Council in collaboration with the National Center for Electronics Recycling and Resource Recycling, Inc., September, 2008; p 3. (43) Osterman, M. We Still Have a Headache with Arrhenius; Electronics Cooling. http://electronics-cooling.com/articles/ 2001/2001_feb_techbrief.php2001, 7 (1), (Accessed 01-02-09). (44) Naveh, A.; Rotem, E.; Mendelson, A.; Gochman, S.; Chabukswar, R.; Krishnan, K.; Kumar, A. Power and thermal management in the Intel core duo; Intel Technol. J. 2006, 10 (2). (45) Silicon Systems, Increasing Flash Solid State Disk Reliability, 2005. http://www.storagesearch.com/siliconsys-art1.html (Accessed 28-10-08). (46) McFedries, P. Microsoft Windows Vista Unleashed; Sams Publishing: Indianapolis, IN, 2007.

ES8020638

VOL. 43, NO. 7, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2549