TEAM 2 Solar-Powered Multi-Seat Computer Kiosk for Tanzanian Classrooms ECE Facilitator Telecomm Facilitator UDSM Solar Advising Professor UDSM Telecomm Advising Professor Management Web Document Presentation/Lab Telecomm Telecomm UDSM Telecomm Jian Ren Kurt DeMaagd Dominick Chambega Aloys Mvuma Jakub Mazur Josh Wong Ben Kershner Eric Tarkleson Joe Larsen Tor Bjornrud Victor Crallet Request for Proposal – October 13th, 2008 Sponsored By: In Cooperation With: Michigan State University University of Dar es Salaam Executive Summary With the increasing proliferation of affordable, reliable personal computers into the marketplace, there is a great demand to develop affordable personal computers for remote and undeveloped areas. One such potential region is rural East Africa, specifically Tanzania. Before deploying a computer system into such harsh conditions, several obstacles must be overcome, including source of electricity, telecommunications, and the savannah climate. The Lenovo Corporation has tasked this team to develop a solar-powered computer workstation that can accommodate up to eight users. The solution must be robust enough to withstand the harsh environment with as little technical maintenance as possible, yet still be affordable for rural schools. 2 Table of Contents EXECUTIVE SUMMARY 2 TECHNICAL SPECIFICATIONS 4 INTRODUCTION BACKGROUND DESIGN SPECIFICATIONS DESIGN CRITERIA CONCEPTUAL DESIGN PHASE I: POWER ARCHITECTURE PHASE II: SYSTEM ARCHITECTURE PROTOTYPES PHASE III: POWER MANAGEMENT PHASE IV: CONTENT 4 4 5 5 6 7 8 14 14 PROJECT MANAGEMENT 15 DESIGN TEAMS AND ROLES 15 REFERENCES 16 IMAGES NOMENCLATURE 16 16 3 Technical Specifications Introduction The primary goal of this project is to help promote education in developing countries by providing grade schools with electronic resources. There are a variety of other groups that have already initiated solutions to this problem. The most prominent group is the One Laptop Per Child Association (hereinafter referred to as OLPC), which has created a cheap, durable laptop known as the XO-1. Other groups such as the Center for Scientific Computing and Free Software (hereinafter referred to as C3SL) have made significant strides in reusing older computers for schools; however, both of those programs have some significant drawbacks. Background The primary competitor identified is the OLPC. The OLPC Association is dedicated to producing low cost laptops and distributing them to low-income areas. There exist several problems with the program, including the per-deployment cost and deployment. The original intent was to deliver a laptop to every child for a cost of $100 per device. The program, however, is unable to deliver the laptop at the $100 target; in fact, the cost to donate a system is almost $200. Deployments also require a minimum commitment of 100 laptops. This represents a very significant financial burden, though once deployed, the XO-1’s are extremely rugged PCs and do not depend on any external power sources. Once deployed, it is difficult to integrate multiple PCs into a cohesive learning environment, and this takes away from educating the students. C3SL’s solution integrates into school systems better, and was widely deployed in the Paraná Digital project. This project involved having the multiple terminals running off of a single computer in multiple schools. This program has been very successful and shows great promise, but there is a critical flaw. The program is entirely software, and this software was intended to run in a classroom equipped with at a minimum basic utilities, such as power and internetconnectivity. 4 Our solution is to integrate the OLPC's ruggedness and the C3SL's novel software solution into one robust package. The design team preceding ours built a solar powered computer system that can be deployed in a relatively durable building. They assembled a solar panel, battery, and a charge controller into a selfcontained solution, such that deployment in a wide variety of climates and locales is possible, but they were unable to decide on the computer system. Our primary goal for this project of integrating the work of our predecessors with a computer system that is suitable for educating youth, regardless of regional or socio-economic boundaries. Design Specifications The core of the design is a single computer powering multiple dumb terminals. There are many ways to create a dumb terminal; these will be discussed later in the proposal. The entire system is connected to an AC/DC inverter, which is powered by a large, deep-cycle battery. The battery is charged via a photovoltaic panel. There is independent monitoring circuitry to ensure the system is functioning properly, which can gather data to recommend ways of improving system performance as well. Once the prototype is complete, we will install it in a school in Tanzania. Lenovo will also be able to mass-produce the system and package it for sale. A variety of organizations, such as governments or humanitarian groups, can then purchase a base station and add any number of terminals. Given that each station functions independent of a power source or communications source, it can be shipped to any location and quickly be installed. Once the system is set up, it will require minimal maintenance, and limited software support will be provided over the Internet. Design Criteria The following requirements are established to decide the feasibility and rating the desirability of the conceptual designs: Stability/Reliability 5 o The system is to operate in a remote area with as little maintenance as possible. Power Consumption o Solar power is the single power source for the system therefore minimal power consumption is a priority. Construction Difficulty o The team has a limited time frame to complete the project and have it packaged ready for deployment. Lenovo Hardware o Implementing the sponsor’s hardware into the system will help keep costs down. Cost o The system is to be implemented in schools with a very limited budget, the lower the cost the greater the chances of system deployment. The criteria (specifications) to be used in the matrices for deciding the feasibility and rating the desirability of the conceptual designs are still being developed, and at least one conceptual design. Conceptual Design The conceptual design for this project is split into four phases. The first is power system design, which for the most part was completed by the previous semester’s team, but was still reviewed by our team. The next phase is system architecture, i.e. how the computers and workstations are set up. After that, we covered power management, and finally, content. 6 Phase I: Power Architecture Figure 1: Power architecture flowchart. Given that the power architecture was in place when the team received the project, and that the schedule and budget are limited, we decided to leave it in its current configuration. A meeting was convened and possible improvements to the architecture were discussed, which could be considered for the production model. Starting from the top down is the solar panel. There are two qualities to consider: efficiency and price. The panel chosen should provide the highest wattage per dollar spent, giving the greatest value. Several cheap, low efficiency panels would be preferable to a single high efficiency panel if they provide a higher wattage per dollar. They would also be more robust; e.g. a single panel could fail and the system would loose a portion of its power generation capabilities, rather than its sole source of power. The panel currently used is the Kyocera KC85TS 85W panel, which operates at 16% efficiency (see Figure 2 for value). A mid-range standard charge controller (CirKit SCC3) is used to regulate the voltage from the solar panel to the battery. The standard charge controller could be improved by replacing it with a maximum power point tracker (MPPT), which is more capable of handling the surplus voltage (> 15V) generated by the solar panel in high irradiance conditions (i.e. direct sunlight). 7 The battery purchased for the project is a 225 Amp-Hour marine deep-cycle battery, chosen for its large capacity and ability to delivery current at a constant voltage for an extended period of time. As in the solar panel, the capacity and price are the two key qualities in consideration (see Figure 2 for value), the life cycle is also very important. It tends to not vary throughout the industry with deep-cycle gel batteries intended for solar use, and therefore did not garner much consideration. This entire system feeds a 1750W power inverter, which ideally operates at 90% efficiency. From here power can be provided to anything that can operate at 115VAC. This may be an issue depending on the area of deployment; a majority of the world operates at 220-240VAC, thus interfacing other components into the power system (such as cell phone chargers) could prove to be dangerous. Figure 2: Component cost/value table. Component Solar Panel Charge Controller Battery Power Inverter Model Number Kyocera KC85TS CirKit SCC3 Capacity Efficiency Cost Value 85W 16% $468.75 0.181 W/$ N/A N/A $44.95 N/A Deka Domintator 8G8D XPower 1750 Plus 225AH N/A $399.07 0.564 AH/$ 1750W 90% N/A N/A Phase II: System Architecture Prototypes The ECE team considered four ideas for the architecture of the system. During a whiteboard brainstorming session, each prototype was sketched, the pros and cons were weighed, and a cost was estimated, as shown in Figure 3. 8 Figure 3: System architecture advantages, disadvantages, and estimated cost table. Cost Outline: Total Cost: Pros: Cons: Laptops Baseline: - Router with advanced features ($200) Per Seat: - Lenovo IdeaPad S10 Latop ($439) - Mount ($50) Thin Client Baseline: - Server ($500) Multi-User Baseline: - Powerful Server ($800) Per Seat: - Diskless Workstation LTSP 1220PXE Thin Client ($285) - Lenovo L197 Monitor ($239) - Keyboard/ Mouse ($30) Per Seat: - Lenovo L197 Monitor ($239) - Keyboard/ Mouse ($30) - Video card ($30) - Optional Software ($100) Base: $200 Per Seat: $489 Easy, Reliable, Serverless, Redundant, Low Power Consumption Base: $500 Per Seat: $554 Easy, Reliable, Stable, Low Power, COTS Small Screens, Defeats Purpose of Designing a New system as Opposed to Donating Laptops, Security Concerns Relatively Expensive, Lenovo Does not Make a Thin Client 9 Blade Client Baseline: - Server ($500) Per Seat: - Small Motherboard with RAM & CPU ($100) - DC-DC Power Supply ($50) - Keyboard/ Mouse ($30) - Lenovo L197 Monitor ($239) Base: $800 Base: $500 Per Seat: $399 Per Seat: $419 Cheap, Lowest Possibly Power Cheaper than Consumption, Thin Single Point of Client, 100% Maintenance, Lenovo 100% Lenovo Hardware Hardware COTS Software is Expensive and Open-Source is Immature, Reliability is Main Concern Lots of Enclosure Work, Reliability Laptop Architecture Figure 4: Laptop system architecture mock-up. The first prototype was a simple laptop-server setup. Each workstation would consist of small laptop (a 10” form factor, such as the Lenovo S10). The laptops would be connected to the Internet either by an Ethernet cable, or even WiFi. Laptops could be run without being directly connected to AC power; a charging station would be setup by the server. This style of architecture would be very simple to configure. The server and the laptops would all be off-the-shelf Lenovo products. The workstations would have low power consumption, given the fact that the monitor, computer, keyboard, and mouse are all combined into one device. Should a laptop be damaged, it would also be very easy to replace, requiring little re-configuration, and no custom engineering. Ultimately, Lenovo has specified that it does not want to simply hand out laptops. The laptops pose a security risk, given that they have value on the open market. Their portability also adds to the security risk. Their all-in-one design also makes them much harder to repair. 10 Multi-Seat Architecture Figure 5: Multi-user system architecture mock-up. One of the most attractive architectures is the Linux multi-seat. Based on some very interesting test cases found, it’s implemented by building a central PC with multiple video cards (4-8), multiple keyboards, and multiple mice. Each workstation would be plugged directly into the server, with individual login names created. . There would be a very low cost for such a setup. No thin clients would be required, only a monitor, keyboard, and mouse. The power requirements would also be lower, given that the CPU and all of its resources would be shared by all of the users. The system would also respond much quicker than a thin client, without the LAN bandwidth and latency issues. There are many websites dedicated to the subject, and the various open source solutions. Unfortunately, these options are buggy and unstable, at best. A for-profit company, Userful, has also popped up, offering a Linux-based closed source solution that is much more reliable than any of the open source solutions found. Trial versions of their software were found to be very user friendly, if not somewhat prohibitive. The largest obstacle was the license, at $100 per seat per year. 11 If the software end of this solution were more mature, Team 2 would highly recommend this architecture, however, given the reliability needed, it would be unwise to implement. Given work by some computer science students, and the open source community as a whole, this could develop into the most robust and costeffective architecture. Blade-Client Architecture Figure 6: Blade-client system architecture mock-up. Another type of architecture discussed was the blade-client (or DIY thin client). For all intents and purposes, it is a homemade thin client. Each workstation consists of a small motherboard (mini ITX), with the accompanying RAM and CPU, but lacking a hard drive. These would be placed inside of a custom enclosure, and connected to a small monitor (17” or smaller), keyboard, and mouse. Each workstation would utilize PXE boot to connect as a thin client to a central server. The entire system could be built using Lenovo components. Custom enclosures would have to be built for each workstation, but the cost of a motherboard, CPU, RAM, and enclosure would be significantly less than a thirdparty thin client (especially when taking into consideration Lenovo’s cost vs. market cost). 12 Unfortunately, this system would be the most difficult to implement, given its the highly customized nature. It would also have slightly higher power requirements than the traditional, depending on the motherboard and CPU used. With the high level of customization, it also increases the opportunities for failures while reducing the reliability, however, Team 2 suggests that Lenovo examine what resources it would take to build its own production quality thin client. Thin Client Architecture Figure 7: Thin client system architecture mock-up. 13 Phase III: Power Management A power management system will be designed to monitor voltage and current levels at key points in the system and use this information to calculate real-time data such as percent battery remaining, time until dead battery, and current charging conditions. To implement this system we will use a PIC 18F series microcontroller for all processing functions. Various voltages can be read using the analog inputs. For current sensing we are using LEM FHS-40P Hall effect sensors. These sensors measure the electromagnetic field created from the current flowing through the wire and convert this to a voltage that the PIC can then calculate the current with. The PIC will communicate with the server using the serial data bus. This is preferred over USB because it is easier to implement and is more consistent from platform to platform. A serial to USB converter would be used if the server lacks a serial port. The power management system will function with or without the server. An LCD screen will display pertinent information. More information can be accessed via several buttons that will allow scrolling through a menu. Several LEDs will display critical information such as power on, low power, and service needed. When power becomes critically low (less than 20% charge remaining in the battery) the PIC will initiate a shutdown sequence which will save important data to the server and then turn off all of the components safely. Once the system has been charged to an appropriate level the system will perform as usual. Phase IV: Content With the majority of potential customers being rural persons that often live in secluded tribes, the content available on the machines must be carefully controlled in order to protect the identity of the tribe. The Internet will be white listed, that is, only a select few websites will be available. Also, other education content, such as digital encyclopedias must be selected to avoid as much Western bias as possible. The Telecommunications teams, both at Michigan State University and the University of Dar es Salaam, have been tasked with seeking out appropriate 14 resources to deliver, without compromising the youth by giving access to restricted content. The Engineering team will also develop software to assist in integrating the machines into the classroom, by offering an electronic hand-in. Students can work on papers on any machine, and then “hand them in” by submitting them to the central server using our software. The instructor can then review them digitally, without having to print them off (infeasible) or collect them manually via USB thumb drive. Project Management Design Teams and Roles Given the size and importance of this project, Team 2 was put together in three distinct teams: Engineering (Michigan State University) Member Jian Ren Jakub Mazur Josh Wong Ben Kershner Eric Tarkleson Role Facilitator Management Web Documents Presentations/Lab Telecommunications (Michigan State University) Member Kurt DeMaagd Joe Larsen Tor Bjornrud Role Facilitator Telecommunications Telecommunications Tanzania (University of Dar es Salaam) Member Dominick Chambega Aloys Mvuma Victor Claret Role Solar Advising Professor Telecommunications Advising Professor Telecommunications 15 References Images All images on the coversheet were obtained from Wikipedia: The Lenovo logo is owned by the Lenovo Group and is being used under the fair-use rationale. The Michigan State University logo is owned by the Michigan State University Board of Trustees and is being used under the fair-use rationale. The University of Dar es Salaam logo is owned by the University of Dar es Salaam and is being used under the fair-use rationale. Nomenclature C3SL – Center for Scientific Computing and Free Software. CPU – Central Processing Unit, refers to the main processor chip on a computer motherboard, not the computer as a whole. COTS – Commercial Off The Shelf, describes hardware or software that may be purchased rather than designed and built. MPPT – Maximum Power Point Tracker, a style of solar charge controller. multi-seat – A type of system architecture in which many workstations are built onto a single machine. PIC – The company that produces the microcontroller used, may also refer to the microcontroller itself. PV – Photo-Voltaic, i.e. solar panel. PXE – Pre-boot eXecution Environment, a manner of booting computers over the network without a locally installed operating system. OLPC – One Laptop Per Child. OSS – Open Source Software. system architecture – The term used within this document to describe how the style in which the workstations are deployed. thin client – A client computer that relies on a central server for a majority of its processing tasks. 16