New User Special Price Expires in

Let's log you in.

Sign in with Facebook


Don't have a StudySoup account? Create one here!


Create a StudySoup account

Be part of our community, it's free to join!

Sign up with Facebook


Create your account
By creating an account you agree to StudySoup's terms and conditions and privacy policy

Already have a StudySoup account? Login here

Topics in Computer Systems

by: Allie West II

Topics in Computer Systems CSCI 7143

Allie West II

GPA 3.51

Richard Han

Almost Ready


These notes were just uploaded, and will be ready to view shortly.

Purchase these notes here, or revisit this page.

Either way, we'll remind you when they're ready :)

Preview These Notes for FREE

Get a free preview of these Notes, just enter your email below.

Unlock Preview
Unlock Preview

Preview these materials now for free

Why put in your email? Get access to more of this material and other relevant free materials for your school

View Preview

About this Document

Richard Han
Class Notes
25 ?




Popular in Course

Popular in ComputerScienence

This 7 page Class Notes was uploaded by Allie West II on Thursday October 29, 2015. The Class Notes belongs to CSCI 7143 at University of Colorado at Boulder taught by Richard Han in Fall. Since its upload, it has received 17 views. For similar materials see /class/231985/csci-7143-university-of-colorado-at-boulder in ComputerScienence at University of Colorado at Boulder.


Reviews for Topics in Computer Systems


Report this Material


What is Karma?


Karma is the currency of StudySoup.

You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!

Date Created: 10/29/15
3 Fred Douglis and Brian Marsh Low power disk management for mobile computers Technical Report MITLiTR753793 Matsushita Information Technology Laboratory April 1993 4 DP Hembold DE Long and BSherrod A Dynamic Disk Spinidown Technique for Mobile Computing In Proc MOBICOM 3996 November 1996 5 T Imielinski S Viswanathan and ER Badrii nath Energy ef cient indexing on air In Pm ceedings ACM SI GM OD 1994 6 Ravi Jain and John Werth Airdisks and air raidmodeling and scheduling periodic wireless data broadcast Computer Architecture News 234223728 September 1995 7 Kester Li Roger Kumpf Paul Horton and Tom Anderson A quantative analysis of disk drive power management in portable computers In Proceedings 1994 Winter USENIX pages 2797 291 San Franscisco California Winter 1994 8 Brian Marsh Systems issues for mobile compute ing Technical Report MITLiTR750793 Matsushi ita Information Technology Laboratory Feburary 1993 9 ns 7 LBNL Network Simulator httpwwwi nrgeelblgovns 1996 10 W M Smith and PS Chang A Low Power Median Access Control Protocol for Portable MultiiMedia Systems In Proc Third Workshop on Mobile Multimedia Communications lIoMuCi3 Sept 1996 11 Stanley Zdonik Michael Franklin Rafael Alonso and Swarup Acharya Are quotDisks in the Air just pie in the sky In IEEE Workshop on Mobile Computing Systems and Applications pages 12719 Santa Cruz California December 1994 Mark Stemm ACM S 1995 IEEE S 95 is a PhD student in Computer Science at the University of California at Ber keley His research interests are in the areas of wireless and wideiarea computer networks mobile computing and oper7 ating systems He received his 13 S degree in Computer Sci ence with University and Departmental honors from Carnegie Mellon University in 1994 and his MS degree in Computer Science from the University of California at Ber keley in 1996 On the WWW his URL is httpwwwcsberkeleyedu stemm His email address is stemmcsberkeleyedu Professor Randy H Katz is a leading researcher in com puter system design and implementation His research expe7 rience has spanned numerous disciplines He has written over 120 technical publications on CAD database manage ment multiprocessor architectures high performance store age systems and video server architectures He was responsible for developing the concept of Redundant Arrays of Inexpensive Disks RAID now a 3 billion industry seg7 ment Katz s recent research has focused on wireless come munications and mobile computing applications From January 1993 through December 1994 Katz was a program manager and deputy director of the Computing Sys7 tems Technology Of ce of ARPA He was responsible for quotwiring the White House to the Internet and also assisted the Clinton Administration in formulating policies related to the National Information Infrastructure and wireless technoli ogies He is a member of the ACM and a senior member of the IEEE 915MHz WaVELAN 2 AGHz WaVELAN Page Response Time sec 100 200 300 400 Attention Span sec Figure 9 Response time as a function of attention span for 915 Mhz and 24 Ghz Wavelan co m Metricom Page Response Time sec 0 100 200 300 400 500 Attention Span sec Figure 10 Response time as a function of attention span for Metricom that increasing attention spans lead to signi cant reductions in energy costs with no userivisible increase in latency Figure 10 shows the response time as a function of attention span for the Metricom NI This illustrates the effect of a large sleep7gtwakeup transition For shorter attention spans the 5 second delay as the interface is powered on has a user visible latency For larger attention spans however the latency to retrieve the web page dominates 6 ConclusionsRecommendations Our measurements ofPDA and Network Interface power and energy consumptions show that Network Interfaces consume a signi cant fraction of the total power on a PDA Additional measurements for sending and receiving packets of various sizes indicate that the power consumed when the interface is on and idle is virtually identical to the cost of receiving pack ets For some interfaces the cost of sending packets can be signi cant when compared to the cost of being idle but application and transportilevel considerations make the idle cost the dominant cost Although the choice of transport layer can have a signi cant impact on the number of packets sent and received by the mobile device the actual power difference is minimal This is because the energy consumed simply by keeping the net work interface on during the transfer contributes the most to the nal energy cost In the presence of a high packet error rate however current TCP sender implementations overrei act to packet losses mistaking them for congestion This slows down the transfer rate which increases the amount of time that the transfer takes and the amount of energy con sumption by the network interface Simulations show that for email our optimizations can reduce the energy consumption to the minimum possible the energy required to receive messages For web browsing fast sleepiidle transitions allow signi cant power savings with no impact on userivisible latency Even for interfaces with longer sleepiidle transitions however signi cant power save ings can be achieved with less aggressive management of the network interface 61 Recommendations for Future Networks Interfaces and Protocols Current generation transport and linkilevel protocols may need some tuning to minimize the power cost of network interfaces Any protocol that leaves a mobile receiver idle unnecessarily such as TCP s backoff in the presence of wireless losses wastes power Even when the protocol is performing correctly inefficient linkilayer scheduling may be the problem a link layer that allocates 2 Mb on a conteni tion basis for 10 mobiles causes each of them to consume 10 times as much power 100 times as much power total as a base station that uses a TDMA scheme to coordinate deliv7 ery of data to receivers Recent work has proposed more intelligent linkilayer schemes to handle this problem 10 The valuable lesson is that network interfaces can consume a significant fraction of the power budget of PDAs and this requires smart software and applications to make sure that battery lifetime is not needlessly shortened 7 Acknowledgments Thanks to Paul Gauthier and Daishi Harada who worked on earlier versions of this work Thanks also go to Bruce Mah from UC Berkeley for providing us with the WWW traces This work is supported by DARPA contract DAABO77957C7 D154 and grants from the California MICRO Program Hughes Aircraft Corporation Metricom and ATampT References 1 Fred Douglis Frans Kaashoek Brian Marsh Ramon Caceres Kai Li and Joshua Tauber Store age alternatives for mobile computers In Pmc 1994 Symposium on Operating Systems Design and Implementation OSDI November 1994 2 Fred Douglis P Krishnan and Brian Marsh Thwarting the powerihungry disk In Pmceedr ings 1994 Winter USENIXSympasium 1994 Metricom Mean power consumption mWsec 200 300 400 5 Wakdsleep periodicity sec Figure 6 Energy vs attention span for Metricom Metricom Mean ServerePDA packet lag sec 0 I I I I I 200 300 400 500 600 Wakesleep periodicity sec Figure 7 Staleness vs Attention Span for Metricom for the user population and the average quotstalenessquot the lag between the time the mail message enters the mail spool and the time that the PDA discovers that the message has arrived Figure 6 shows the average energy consumption as a funce tion of the attention spanAs the attention span increases the energy consumption decreases Figure 7 shows the corre sponding quotstalenessquot which increases linearly as a function of the attention span The results are quite promising with an approximate staleness of two minutes the power con sumption drops by 20 This attention span reduces the energy consumption to the cost of retrieving the email mes sages 5 Web Access Simulation In this section we describe optimizations that can be used to reduce energy consumption for Web browsing applications We brie y describe the data trace collection below 51 Trace Collection and Processing We used traces of HTTP traffic at UC Berkeley as input to a simulator which experimented with different power savings strategies For each workstation we kept track of the start times and transfer sizes for each outstanding HTTP connece 15000 10000 915MHz WaveLAN 2 40112 WaveLAN 5000 Energy per Page mWAsec 100 200 300 400 500 Attention Span sec Figure 8 Energy per page as a function of Attention pan for Wavelan NIs tion For each user we divided time into work when at least 1 outstanding connection was outstanding and think when no connections are active phases These posteprocessed traces form the input to the simulation 52 Power Saving Strategy The power saving strategy evaluated in this section attempts to reduce effective power consumption during the think time portions of the traces We turn off the network interface after the user has been in a think phase for more than a certain amount of time called the attention span It stays in that state until the user sends data in this case a HTTP request from the interface It is important however to distinguish between large think times and times where the user has stopped using the application For our measurements we speci ed a maximum attention span of 5 minutes after which we considered the PDA to have been turned off Any think times more than 5 minutes were excluded in the simue lation In this way we do not falsely claim energy savings when the user would have simply turned of the device imme diately 53 Simulation Setup and Results The web simulation uses the N1 and transportelevel measure ments from Section 2 and Section 3 as well as the traces described above The outputs of the simulation are two metrics of perfore mance Average energy cost in mWeseconds of an HTTP page retrieval The average latency for the initiation of a Web page access This measures the average amount of time to complete the first HTTP request ofa work phase with the assumption that Web page accesses are a single html document followed by a number of inline images Figure 8 Figure 9 and Figure 10 show the simulation results for the Wavelan and Metricom devices Figure 8 shows the energy per page as a function of the attention span for the Wavelan le and Figure 9 shows the response time as a function of attention span For the Wavelan Nl we can see Power ounsummmn utvanuus nanspun PmtuculSSendRecv Power Ontv UDP rm StzeWmduW r UDP Unhmtted StzeWmduW 7 TOP TOP Wllh Delayed Acks 44 2 Power Consumpunmwrsemnus onnnnn ennnnn annnnn we Transtev StzEOWtes znnnnn Figure 3 Energy for different transport protocols only including SendRecv Power ounsumpn annn UDP rm StzethduW e UDP Unhmtted StzethduW 7 TOP 7 mun TOP Wllh DElaved ennn annn t annn F nwerCnnsummmmmvv semnds zann mun Inclnnn tnnbnn Budnnn Budnnn we Transtev StZEOJVtes Figure 4 Energy for different transport protocols including SendRecv and Idle fer time comes into play Because the amount of time that the receiver waits for packets from the sender to arrive is much larger than the relatively small amount of time that the receiver actually sends or receives packets the idle cost dominates the cost to send or receive packets and the differ ence between the transport protocols is eliminated 34 The Effect of Error Rate on Energy Section 33 shows that Idle makes the greatest contribution to nal energy cost and that for low error rates the different transport protocols behave similarly Figure 5 shows the effect of a higher error rate on energy consumption In the presence of a high packet error rate the difference is more significant As shown in BSAK95 TCP mistakes packet losses for congestion and reduces the transmission rate From a power standpoint this decreases the value of Band increases the total energy cost A more intelligent scheme that does not mistake wireless packet losses for congestion would not have this problem In the following sections we use the results from the trans port level simulation to experiment with applicationispecific policies for reducing energy consumption of network inter faces Pav evcansu mum mm StzethdaW 4 rev 7 15mm touuu zuuu uuuu auuu suuu mm M zuuu pmrcmsummmrmwsemms 4e zuuuuu mm summon Euuuuu eons Yvanstev SVIEthes Figure 5 The effect of wireless losses on energy con sumption From our transportilevel measurements we have learned that The dominant energy cost of any transport protocol is not the number of packets sent or received but the amount of time that the transfer takes to complete This property means that the energy cost can increase signifi7 cantly in the presence ofwireless losses where a receiver must wait for a TCP sender to recover from packet losses The results from our transportilevel measurements are used in the applicationispecific experiments of Section 4 and Section 4 Mail Simulation In this section we describe applicationispecific optimizai tions that can be used to reduce the energy compositions of network interfaces while using electronic mail applications We start with a brief description of the trace data used for the experiments 41 Data Collection We used the user population of the Computer Science Divii sion at UC Berkeley to measure mail activity The arrival times and sizes of mail messages appearing in the Division mail spool was collected This trace was used as a sample workload to the simulations of Section 42 42 Simulation Setup and Results In our strategy for reducing energy consumption the PDA wakes up periodically bringing its Nl from a sleep to idle state and checks for new mail Like approaches in 5 6 and l l the availability of new mail is broadcast periodii cally so the FDA does not have actually transmit any packets to check for mail We define the attention span as the amount of time that the PDA waits before waking up and checking for new mail We ran the simulation for attention spans ranging from 60 sec onds 1 minute to 600 seconds 10 minutes in 15 second increments and measured the average energy consumption PuWevCunsumptmn Fuvvanuus Packet Stzes Memcum Send k m Memcum Recv Memcuml Power Consummmmwrsemnus thn z m a m u m lm elm mu Elm a m lm 11 n Packet StZEOJVtes Figure 2 Energy consumption for different packet sizes for Metricom mately double that of idling for the same amount of time This would imply that sending is much more expensive than being idle and network protocols should minimize the num7 ber of packets sent As we will see in Section 3 however other transportilevel considerations have a more significant impact on the energy cost In addition for the applications of Section 4 and Section 5 the amount of time that the NI spends sending short acknowledgments is outweighed by the time spent receiving data packets We believe that these applications or similar ones where the PDA is retrieving rather than sending large amounts of data will be the most common applications on future PDAs From these measurements we may conclude that 1 Receiving packets only costs slightly more than the idle cost 2 Sending packet costs more than receiving and can be signi 7 cant when compared to the cost of being idle but only if the mobile is sending large amounts of data to the wired network 3 Transport Layer Simulation In this section we examine different transportilevel protoi cols and find the energy costs to send data to a mobile receiver for each transport protocol We start with a simple breakdown of transportilevel energy consumption 31 Breakdown of Transport layer Energy Consumption We can break down the energy consumed to complete a bulk transfer of b bytes as follows for a xed data packet size and a xed acknowledgment size Idle 1 Energy SendRecV Idle SendRecV aE dE a d Where a is the number of acknowledgments sent Ea is the energy cost to send a single acknowledgment dis the num7 ber of data packets sent Ed is the energy cost to send a single data packet is the instantaneous idle power and Bis the effective bandwidth of the transfer Our goal is to see how these two components of the energy cost change as the choice of transport protocol changes In all of these simula7 tions we assume a transfer from a fixed source to a mobile receiver We compared four different transport layer protocols in terms of the number of acknowledgment packets they gener7 ate the number of packets that they send to the mobile device and the amount of time necessary to accomplish the transfer These were 1 TCP Reno Using this protocol the receiver generates an acknowledgment for every data packet sent 2 TCP Reno with delayed acknowledgments Using this pro tocol the receiver generates an acknowledgment for every other data packet 3 Reliable UDP fixedisize window Instead of depending on acknowledgments for ow control this protocol uses rate con trol in combination with a xed size error recovery window of size W We used a window size of 10 Each window is acknowli edged by the sender with a single selective acknowledgment and any missing packets in the window are retransmitted by the sender The receiver sends on average a little more than one acknowledgment for each w packets 4 Reliable UDP unlimited window This is a special case of the above UDP scheme when the ow control window is equal to the number ofpackets sent The primary difference between these schemes is in the number of acknowledgments sent by the mobile device and the number of times that a duplicate packet will be received by the sender 32 Methodology The scenario we used was a three node network including a source base station and receiver The source and base stai tion were connected with a high bandwidth low error rate link and the base station and mobile were connected with a lower bandwidth higher error rate link We simulated the TCP protocols using the Network Simulator ns 9 We simu7 lated the Reliable UDP schemes by deriving formulas that showed the number of packets sent and received for a given bulk transfer size and packet error rate To compare the prof tocols we kept track of the total length of the transfer and the values of a and d We then used the information extracted from the data in Figure l and Figure 2 to generate the energy drain for each packet sent and received as well as the energy cost for the entire transfer 33 Simulation Results Figure 3 shows the contribution that SendReCVmakes to the energy cost for a variety of transfer sizes for the 915 Mhz Wavelan The x axis shows the transfer size and the y axis shows the energy cost in mWiseconds These results show that the UDP protocols which send fewer acknowledgments use less energy When the contribution from Idle to the total energy cost is included however Figure 4 the total trans is a diffuse Infrared PCMCIA interface with a range of approximately 5m and a userevisible bandwidth of approxi mately 850 kbitssec The Apple Newton and Sony Magic link are commercially available PDAs To measure power consumption of the PDAs we measured the devices while performing tasks designed to stress one subsystem of the device for example pen input speaker output etc and then averaged the measurements to obtain typical behaviorl We measured the network interfaces while quotidlingquot powered on but not sending or receiving packets sleeping powered off but still connected to the device and sending and receive ing packets ofvarious sizes We also measured the quotwakeupquot time defined as the amount of time from when the device was brought out of its sleep state until the time that the rst packet can be sent 21 Methodology To measure the power consumption for steady state behave iors we required both current and voltage measurements We used a digital oscilloscope to measure the voltage and current draw of the various devices The current draw was actually measured by using a small resistor and measuring the voltage drop across the resistor For the PDAs and Rico chet modem which have their own external batteries we measured the voltage and current at the battery terminals Although the Ricochet Modem currently has its own bate tery Metricom has plans to make a Ricochet Modem using a PCMCIA form factor For the PCMCIA NIs we measured at the power pins coming into the card For instantaneous operations such as packet transmission and reception we made several measurements and averaged these together to obtain an average value The digital oscilloscope produced bitmaps of the instantaneous voltage across the resistor over time We posteprocessed the bitmaps to obtain the area under the curve energy We also verified that the voltage drops while taking measurements were not large enough to bias our results 22 Measurement Results Table 1 shows the average power consumption of the two PDAs and the Network Interfaces An entry of quot7 means that the device was not measured in that state and an entry of quotNA means that the state is not applicable to the device The Metricom modem has a unique quotwakeupquot state when the Metricom modem turns on it registers with the network and consumes more power for approximately the rst minute of activity One important observation is that for all possible combinations of network interfaces and PDAs the power consumed by the NI is comparable to or even more than the power consumed by the PDA This is a clear indication that power management of the NI is essential Also notice that the Metricom device has a much longer wakeup time than the other NIs This will affect the usefulness of some of the 1 More detailed measurements of the devices can be found at htthwwwcsberkeleyedustemmpowerhtml Device Sleep Idle Wakeup PowermVV Wakeup Time ms Power mVV Wavelan 1773 13189 100 915 Mhz Wavelan 1430 11486 100 24 Ghz Metricom 935 346943103 5000 IBM IR 7 3496 100 Newton 1642 11878 NA FDA Magic 31203 700 NA Link FDA Typical 7 8000 Laptop TABL E 1 Power Consumption for network interfaces and devices Puwev ounsummmn Fm Venous Packet Sizes mu Inn 3cm mu 5mm BUD mu m Bun mun M n Packet SizEbvt25 Figure 1 Energy consumption for different packet sizes for 915 MHz Wavelan optimizations described in Section 5 Once we had the measurements of the NIs while in their idle and sleep states we performed more detailed measurements of the network interfaces to determine if the energy con sumption differed significantly as the interface sent and received packets of various sizes Figure l and Figure 2 show the results of these measurements for the 915 Mhz Wavelan and Metricom devices respectively The xeaxis shows the packet size in bytes and the yeaxis shows the energy consumption in milliwatteseconds There are two lines that correspond to sending and receiving packets Also included is a quotbaselinequot measurement that indicates how much energy is consumed by keeping the interface on and idle for the same amount of time that it takes to send a packet One obvious feature in the graph is that receiving packets only costs marginally more energy than being idle This is also true for sending packets on the 915 Mhz Wave lan For the Metricom device the cost of sending is approxi Measuring and Reducing Energy Consumption of Network Interfaces in Hand Held Devices Mark Stemm and Randy H Katz stemmrandy CSBerkeleyEDU Computer Science Division Departinth of Electrical Engineering and Computer Science University of California at Berkeley Berkeley CA 94 72071 7 7 6 Abstract Next generation handheld devices must provide seamless connece tivity while obeying stringent power and size constrains In this paper we examine this issue from the point ofview of the Network Interface NI We measure the power usage of two PDAs the Apple Newton Messagepad and Sony Magic Link and four NIs the Metricom Ricochet Wireless Modem the ATampT Wavelan operating at 915 MHz and 24 GHz and the IBM Infrared Wireless LAN Adapter These measurements clearly indicate that the power drained by the network interface constitutes a large fraction of the total power used by the PDA We then examine two classes of optie mizations that can be used to reduce network interface energy con sumption on these devices transportelevel strategies and applicationelevel strategies Simulation experiments of transport level strategies show that the dominant cost comes not from the number of packets sent or received by a particular transport protoe col but the amount oftime that the NI is in an active but idle state Simulation experiments of applicationelevel strategies that signi cant energy savings can be made with a minimum of userevisible latency 1 Introduction Handeheld devices coupled with wireless network interfaces are emerging as a new way to achieve seamless connectivity However these new devices have power cost weight and size constraints that are more stringent than most laptop computers The goal of achieving seamless connectivity while staying within limited size and energy constraints is challenged by the addition of a large power consumer to a personal digital assistant FDA a wireless network interface NI Current wireless network interfaces consume as much power as an idle PDA For example the network interfaces we measured consumed from 350mW to 1300mW when idle and the PDAs we measured consumed from 700mW to 1200mW when idle Although much work has been done in reducing the power consumption of other peripheral devices such as disks in laptop computers 3 2 l 7 Li94 8 4 little work has been done to reduce NI power consump tion in handheld devices This paper presents detailed measurements of the power and energy consumption of several wireless network interfaces to determine the powerenergy drain of devices in their sleep idle packetesend and packetereceive states We then examine two classes of optimizations that can be used to minimize energy consumption of wireless network interfaces trans port level optimizations and application level optimizations For transport level optimizations we examine different choices of transport layer protocols using simulations to examine the relative power tradeeoffs when sending equal amounts of data from a wired sender to a mobile receiver We nd that the dominant cost in the energy usage of a trans port protocol is the time that the transfer takes to complete not the number of packets sent or received by a particular transport protocol For applicationelevel optimizations we focus on two appli cations that we expect to be the killer apps for PDAs electronic mail and web access We use realeworld traces combined with simulations to experiment with application speci c energy savings strategies Results show that signifi cant energy savings can be made with a minimum of user perceivable latency In particular for electronic mail applica tions the energy consumption can be reduced to the mini mum amount the energy required to retrieve a piece of electronic mail For webebrowsing applications energy con sumption can be reduced by a factor of four with virtually no impact on userevisible latency The rest of this paper is organized as follows Section 2 pre sents our power measurements of the NIs and PDAs Section 3 presents our transportelevel optimizations designed to reduce NI energy consumption Section 4 presents our applicationespeci c policies for email applications Section 5 presents our policies for reducing energy consumption while web browsing and Section 6 presents conclusions and rec ommendations for future protocols and network interfaces 2 Measurements In this section we describe the methodology used to mea sure power consumption of devices and our results for seve eral wireless NIs and PDAs We measured power consumption of two PDAs the Apple Newton Messagepad 100 and the Sony Magic Link PIC 1000 and four network interfaces ATampT s Wavelan PCMA CIA card operating at 915 MHz and 24 GHz Metricom s Ricochet Wireless Modem and IBM s Infrared Wireless LAN card The ATampT Wavelan is a direct sequence spread spectrum PCMCIA interface with a range of approximately 40m and a userevisible bandwidth of 16 Mbits The Rico chet Wireless Modem is a frequency hopping spread spec trum modem device with a range of approximately lkm and a userevisible bandwidth of 50 kbits the IBM Infrared card


Buy Material

Are you sure you want to buy this material for

25 Karma

Buy Material

BOOM! Enjoy Your Free Notes!

We've added these Notes to your profile, click here to view them now.


You're already Subscribed!

Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'

Why people love StudySoup

Steve Martinelli UC Los Angeles

"There's no way I would have passed my Organic Chemistry class this semester without the notes and study guides I got from StudySoup."

Amaris Trozzo George Washington University

"I made $350 in just two days after posting my first study guide."

Jim McGreen Ohio University

"Knowing I can count on the Elite Notetaker in my class allows me to focus on what the professor is saying instead of just scribbling notes the whole time and falling behind."


"Their 'Elite Notetakers' are making over $1,200/month in sales by creating high quality content that helps their classmates in a time of need."

Become an Elite Notetaker and start selling your notes online!

Refund Policy


All subscriptions to StudySoup are paid in full at the time of subscribing. To change your credit card information or to cancel your subscription, go to "Edit Settings". All credit card information will be available there. If you should decide to cancel your subscription, it will continue to be valid until the next payment period, as all payments for the current period were made in advance. For special circumstances, please email


StudySoup has more than 1 million course-specific study resources to help students study smarter. If you’re having trouble finding what you’re looking for, our customer support team can help you find what you need! Feel free to contact them here:

Recurring Subscriptions: If you have canceled your recurring subscription on the day of renewal and have not downloaded any documents, you may request a refund by submitting an email to

Satisfaction Guarantee: If you’re not satisfied with your subscription, you can contact us for further help. Contact must be made within 3 business days of your subscription purchase and your refund request will be subject for review.

Please Note: Refunds can never be provided more than 30 days after the initial purchase date regardless of your activity on the site.