Locally Networked Satellite-Based Computer Labs for Tanzanian Classrooms Final Proposal February 20, 2009 Sponsored By: In Cooperation With: University of Dar es Salaam Michigan State University ECE 480 – Design Team #2 – Spring 2009 Management Webmaster Document Prep. Presentation/ Lab Brian Holt Daniel Newport Steven Sadler Kevin Bishop Executive Summary With the increasing dependency on technology there is a great demand to develop affordable personal computers for remote and undeveloped areas. One potential region is rural East Africa, specifically Tanzania. Before deploying a computer system into such harsh conditions several obstacles must be overcome, including providing a reliable source of electricity to the system, telecommunications, and the savannah climate. The Lenovo Corporation has tasked this team to develop a computer workstation that can accommodate up to eight users. The solution must be robust enough to withstand the harsh environment and also be affordable for rural schools. 1 TABLE OF CONTENTS Technical Section Introduction……………………....Page 3 Background……………………....Page 5 Design Criteria…………………...Page 6 Conceptual Design……………….Page 7 Ranking of Conceptual Design…..Page 10 Project Management Section Project Management Plan………..Page 14 Costs Section Approximate Costs…………..…...Page 15 Fast Diagram/ References Fast Diagram………………..……Page 16 References………………………..Page 16 2 Introduction Many schools in rural Africa are without computers and network access, and have almost no books for their students to use. Having internet access could greatly improve the current situation of these schools. However, many schools are in locations with no electrical service or network connectivity. While we may not be able to purchase books for these schools we are able to provide a means to improve their education via computers and the use of the internet. Lenovo has chosen to support the development of an eight-seat multi-user computer workstation complete with network access. Since installing a computer system in rural Africa will present many issues we will attempt to simulate, and educate ourselves as much as possible on, as many of these issues as possible prior to our arrival in Tanzania. We will begin this semester’s project by reviewing the previous semesters efforts to install a similar system in Tanzania. While there are a few differences between the two projects, such as the previous teams use of solar power, we believe we can use the previous project in order to evaluate what measures they took in order to successfully install their system in Tanzania, issues that they dealt with and had to overcome, and what we can do to improve upon the previous design. The following is the main criteria that we must consider in order to implement a successful project design by May 9th, our scheduled deployment date. The system that we install must be robust and be able to perform under any weather condition. This system must be reliable in heat that is typical in Africa, with an average year round temperature of 85 degrees Fahrenheit and average highs of 110 degrees Fahrenheit. The system must also be able to operate under the most unpredictable of power conditions. The secondary school that we will be installing this system in has access to a power grid, however, that grid is subject to unpredictable power outages ranging in time from a couple of minutes to full days. This power grid is also subject to varying voltages, which could be harmful to the computer system and its components. This system must be as maintenance free as possible. We are installing this system in an area where neither the teachers nor the students have any previous experience with operating a computer. If problems were to occur within the system debugging and fixing these problems may not be possible for large amounts of time. We will be routers equipped with modified antennas to create the gain and signal strength necessary to accomplish the task of internet connectivity between location of the primary school that last semesters team installed a workstation at and this semesters location at the secondary school, which are separated by approximately two miles. We are also designing this system so that it may easily be expanded beyond the two schools that will have network access upon completion of this semester’s project. With 3 the current multi seat system we are designing we will have two computers each equipped with the ability to accommodate up to four users per computer. We are also able to easily expand upon the current network connections due to the fact that the proposed antenna design is capable of sending a usable signal up to five miles. 4 Background In the fall semester of 2008 the Lenovo Corporation made a unique opportunity available to a design team in ECE 480. The task would be to introduce a low cost, low maintenance, solar powered Linux based computer workstation into a primary school in rural Africa capable of seating up to eight users. The team would face many obstacles while preparing to deploy the system including the savannah climate, telecommunication and the lack of a reliable form of electricity within the village in which the school was located. The team was able to use solar panels to charge a small battery bank, which included a custom engineered management system to monitor voltages and currents from the solar panels and the battery. The system was also able to monitor the temperature within the case that stores the batteries and monitoring system, to ensure that the system is shut down if an unsafe temperature is reached within the case. This semester Lenovo Corporation has given our team the task of developing a similar computer workstation, also capable of accommodating up to eight users. This workstation will again be entered into the harsh climate of Tanzania but, contrary to last year’s workstation, will accommodate users in a secondary school, approximately two miles from the primary school location from last semester. However, this semester we will be able to utilize the fact that the secondary school is located on the power grid. While we are fortunate that the secondary school is located on the power grid we are still forced to do with the reoccurring problem of unreliable power since this grid is subject to unpredictable “blackouts” and varying voltages. To conquer the power problems we will incur while attempting to place the workstation in Tanzania we will use an uninterruptible power supply (UPS) system to charge a small battery bank, which will be used to run the workstation when the power from the grid is unavailable. This will allow us to ensure that brief power disruption does not cause injuries, data loss, or system failure. In the event that power is no longer available from the grid the UPS will allow the user to continue normal operation on the workstation for up to four hours. Ideally the UPS will include a monitoring system to warn the user that the battery bank is running low, in order to ensure that important data is saved and the proper shut down procedure is done in the event that power from the grid is not restored. This semester we will also be installing modified antennas to two routers, one that will be placed at the primary school and one at the secondary school. These antennas will allow us to use the satellite router that was previously installed in the primary school location and access the internet at the new location, which is approximately two miles away. We will also be moving the satellite router to the secondary school, in order to release some of the power draw from the solar power system. This project will be completed throughout the course of this semester and will be installed by the members of our team, as well as several members from the previous semesters team, into the Tanzanian secondary school beginning on May 9th, 2009. 5 Design Criteria Low cost (priority: 4) – This is the most important design criteria. Since this system is being designed for Third World countries to purchase, the cost needs to be as low as possible. In particular, the computing solution will be designed to minimize the cost per seat. Also, normal (store bought) wireless routers with modified firmware will be attached to high-gain directional antennas so that a single internet connection can be shared amongst multiple areas (schools). Low maintenance (priority: 3-4) – Also one of the more important criteria. These systems are being designed for areas that will have little to no available support. Thus, they must be designed for robustness and require little maintenance. All materials used for the antenna will be designed to handle a the expected harsh conditions. Furthermore, the computing solution will be user friendly and secure. The uninterruptible power supply (UPS) will help shield the computer from fluctuations in power, and the UPS itself will be designed to handle an unpredictable energy supply. Low power consumption (priority: 1-2) – Low power usage is less of an issue than in the Fall 2008 project, however, a system that achieves this will increase the uptime of the UPS should the power fail. The multi-seat computer system will help make this system use less power by consolidating the required electronics into a single PC. Safety (priority: 3-4) – Safety is a high priority in more than just the physical sense. One of the primary means of accomplishing this is the use of a white list on the computing system to help control the content that school-aged children are allowed to view. The white list will also make the system more maintainable by blocking sites that could harbor malicious code. Also, the development of appropriate content for the system will allow for an enriching yet safe experience. Expandability (priority: 2-3) – While the situation we are designing for is fairly unique, the final system should be expandable so that it is usable as universally as possible. The multi-seat system currently installed has four seats, but could be expanded to eight on the same hardware. Furthermore, if a motherboard with more PCI slots was chosen the system could possibly allow more than eight simultaneous users. The antenna will be designed for a five mile hop so that it is usable in a multitude of situations. 6 Conceptual Designs Multi-seat computing system: There are a variety of options for the multi-seat computing system; one of the leading items of debate is the choice of operating system. Candidates thus far are Ubuntu Linux (used by the Fall 2008 team), Open Solaris, and Red Hat Linux. Initial testing of Open Solaris showed that it was not as user friendly as needed for a system of this nature. In an effort to maintain consistency the Spring 2009 group has again chosen Ubuntu Linux as the candidate operating system. This choice ensures that the environment will be intuitive and easily usable by young children. Also, this allows for the continued use of the Multi-seat Display Manager (MDM) which may not be usable in a non-Debian based operating system. Although less of a concern than those previously mentioned, the use of Ubuntu makes the system easier to set up for an administrator or anyone else responsible for maintenance. A secondary design decision is regarding the hardware used in our proposed design solution. Doing some rough estimation using data compiled by the previous semester's team the cost to implement a thin-client based computing system would be too high to justify switching to it over the current design. Although such a system would be highly scalable, the cost per seat is significantly higher due to the thin client hardware itself and the cost of a powerful central server to process the data. Again, the Spring 2009 team has chosen to move forward using the same hardware as the previous semester. The Lenovo ThinkStation S10 is a powerful workstation PC that more than suits the needs of our target user, and gives us a lot of processing power to run the multi-seat system. We have opted to use two S10 workstations minimally running four seats each for Manyara Secondary School. This should help with stability concerns, but at the same time allows for expansion up to a total of sixteen seats. On site administration will hopefully be easier since each PC will be less complex to set up. Antenna: 7 The primary decision regarding the point to point link antenna is whether to buy a commercially available one or to construct our own. A directional antenna designed for the 2.4GHz band that could achieve our target distance of two miles would cost upwards of $50. The advertised gain of such an antenna was generally around 20 dBi, but sometimes up to 25 dBi (the higher the gain the more costly the antenna). There is no guarantee that a commercially bought antenna could actually reliably transmit a signal that distance, and buying such an antenna would make it unmodifiable (e.g. to make it more rugged to inclement weather). Preliminary research regarding constructing our own antenna led us to a “Do It Yourself” antenna constructed from an aluminum can. The can acts as a waveguide feed, and would be directed into a parabolic dish to achieve even more gain. A prototype dish could be constructed from thick gauge copper wire and aluminum window screen for low cost, however the final design would need to be considerably more sturdy for it to be deployed. Concerns over the sturdiness of this antenna led us to seek advice from various antenna experts. The antenna prototype we are in the process of constructing is a combination Yagi-Quad array constructed from the handle of a hockey stick and other readily available parts. This design makes the antenna more compact than a pure Yagi or Quad, and is anecdotally quoted at a range considerably higher than our target (10 miles). Another attractive feature of the YagiQuad combination is that it is easier to achieve impedance matching with the router which will allow for increased gain. Also, Yagi-Quad arrays generally provide better gain than patch or parabolic dish antennas. Once our prototype has been properly tested we will construct a version that can withstand adverse weather conditions. Uninterruptible Power Supply (UPS): Again, one of the major decisions regarding this aspect of our project is whether to purchase a UPS that is commercially available or to construct our own. Analysis of the components required to construct our own UPS roughly price it at about $900 ($400 for a deep cycle gel-cell battery, $300 for a DC to AC inverter, and $200 for a gel-cell DC charger). This would create a UPS with an uptime slightly longer than our target. However, this design does not take into account an interface to the computer to let the users know how much time is left until the battery is dead. Also, this design does not take into account the wide variation in voltage from the wall outlets on the power grid (180VAC-240VAC), which could cause damage or at least reduce the life of the gel-cell DC charger). Since high reliability and low cost are important factors, this design is most likely out of the scope of this semester's project. Our other option is to purchase a commercially available UPS from a reputable vendor. Such a system will be designed to handle extreme variation in input power, and will also be able to interface directly with our computer system to notify the users when they should shut the system down. Using a proven system will ensure that the system can be shut down properly without losing data, and more importantly will protect the hardware from an improper shut down. One current candidate UPS (built by APC Power Systems) can provide around twenty minutes of uptime using our proposed computer system, and costs around $500. We feel that although this stock solution does not achieve the desired uptime, in terms of both cost and overall system 8 lifespan this is a better option. There is a strong possibility that we could create a battery bank to be hooked into this UPS, thus expanding the available charge capacity significantly. Router: The router will be a simple wireless device obtainable at any retail electronics store. By flashing the firmware of this device with Open WRT we can greatly increase the amount of administrative options that are available for the router to create a secure and reliable networking base for our computing systems. Setting up the router in this manner will allow us to install a filter and whitelist to control the content available to the school children. We can also control the route of the network traffic so that it is only sent to the appropriate destination, which as the network expands will greatly increase efficiency. The router hardware being used is the Linksys WRT54G v8 home wireless router which costs about $40. Low cost and proven compatibility with Open WRT are the leading factors in this decision. Content filtering and traffic control will be achieved using the Squid Caching proxy freely available for Open WRT. Our routers will be retrofitted with custom made antennas to allow for an expandable network in even the most remote of locations. By creating such an inter-networking scheme a single Internet connection can be shared amongst multiple schools to help reduce costs even further. 9 Ranking of Conceptual Design Computer System: Criteria Thin Client Based Multi-seat System System with 8 Monitors Lenovo ThinkStation S1 Server: $1200 Lenovo ThinkStation S1 Server with 4 Graphics Cards: $1300 Lenovo L197 LCD Monitor: $239 each Low Cost (Priority: 4) Keyboard/Mouse: $30 each Expandability/Flexibility (Priority: 2-3) Low Maintenance (Stability) (Priority: 3-4) Performing maintenance may be difficult on the thin client. Lenovo L197 LCD Monitor: $239 each Keyboard/Mouse: $30 each Cost per Seat: ~$594 Cost per Seat: ~$431 Cost per Seat: ~$704 System is quite expandable, but at a higher cost. May need to upgrade the central server (costly) if too many clients are connected. Lenovo ThinkStation S1 Server with 4 Graphics Cards: $1300 Lenovo L197 LCD Monitor: $239 each Keyboard/Mouse: $30 each Diskless Workstation Thin Client: $285 each Two Server Multi-Seat with 4 Monitors Each Possible to expand this system to contain up to 8 Use half as many seats on seats. Could use a each server, allows more powerful system to be expanded in server to get more the future. Slightly more seats than this, still costly. lowest overall cost. Centralized point Two points of of configuration, configuration; however, however, also a they are mirrored so single point of complexity stays the failure. If system same. Even if one PC goes down there goes down there will still will be no be 4 (or possibly more) computers. seats available for Antenna: Criteria Commercially Parabolic Dish with 10 Yagi-Quad Array Low Cost (Priority: 4) Expandability (Priority: 2-3) Low Maintenance (Rugged Design) (Priority: 3-4) Purchased Cost per Antenna: ~$60, could be more dependent on gain (need to do further testing). Also need to include cost for cables to run to the router, should be approximately the same for each antenna (~$15) "Cantenna" Feed Cost per Antenna: $15 (made from an aluminum coffee can available from most stores, thick gauge copper wire, and aluminum window screen scraps) Cost per Antenna: $20 (made from some base material (i.e. hockey stick handle), copper wire for the Quad elements, and threaded rod for the Yagi elements) Unsure if a commercially built antenna will provide us with enough gain to properly transmit our signal. Unsure if gain will be high enough for our needs. Could have issues with constructing a well designed antenna. The gain for such an antenna if well designed is anecdotally quoted at a range well beyond our requirement (10 miles). This should provide for a highly expandable point-topoint mesh network. Materials used to build the antenna may not be sufficient for the climate in Tanzania. May not be able to easily modify without messing up antenna characteristics. Initial prototype materials would most certainly not Initial prototype withstand the should be fairly elements in Tanzania. ruggedly designed. A more rugged design May have to do some would be possible, but minor treatment of the the cost would rise materials to protect significantly. Also, them from the we would likely have climate. Compact to completely footprint should also redesign the final help this design. product in terms of impedance matching. Uninterruptible Power Supply: Criteria Low Cost (Priority: 4) Commercially Purchased UPS APC UPS Cost: ~$500 Battery Cost: ~$500 Additional Battery Cost: $150 DC-to-AC Inverter Cost: $300 11 Self-Built UPS DC Charger Cost (AC Input): $300 Cost could vary dependent upon the power needs for a particular application (e.g. might need a more powerful inverter for a higher overall wattage system) Expandability (Priority: 2-3) Low Maintenance (Priority: 3-4) Safety (Priority: 3-4) The stock UPS supports approximately 20 minutes of The self-built UPS would give uptime using our current a high amount of electronic configuration. expandability. The only However, this UPS supports requirement would be to the installation of additional ensure that the DC charger is batteries to increase uptime. powerful enough to properly This feature should help us charge the battery bank. meet our desired uptime requirement. Using a proven design will The self-built UPS gives no ensure that the UPS functions guarantee that it will properly properly. One of the more handle the varying input important features is an power from the grid (could interface directly to the PC to cause damage to the system or inform users of the at least lower its lifespan). approximate amount of power Also, there is no built in PC left. This will give ample interface, so one would have warning for when to shut to be designed. Failure of this down the system to avoid data component could lead to loss and help protect the damage to the hardware (both hardware from improper UPS and computer system). shutdown. The commercially bought UPS is already in an enclosure so that the electric components A self-built UPS would are not in plain view. This will require a proper enclosure to increase the safety of the ensure that it is safely isolated. system in case anyone attempts to modify it. Proposed Design Solution: Given the above conceptual designs, and our understanding of the system requirements, Team 2 has selected an overall design to move forward with. The basis will be a similar computer system used by the previous semester's team. There will be two ThinkStation S10 12 workstations powering four seats each in the Manyara Secondary School. Each of these workstations will run Ubuntu Linux, and will be configured into a multi-seat system using the Multi-seat Display Manager (MDM). By using two workstations with only four seats each the system will be expandable if desired, and possibly more stable. Four additional seats will be added to the system currently installed in the Baraka Primary School. The satellite Internet link currently installed in the Baraka Primary School will be moved to the Secondary School so that it can be on the power grid. In turn, a wireless link will be created between the two schools using cheap, off-the-shelf wireless routers with modified antennas so that a single Internet connection can be shared. The routers selected are the Linksys WRT54G v8 routers and will be "flashed" with an Open Source firmware so that their administrative capabilities are expanded. The firmware used will be OpenWRT, and will allow us to put the Squid Caching proxy directly on the router to help control available content. A whitelist will be implemented on the router, and will be accessible from any of the terminals using a simple SSH connection (Administrators only). Also, we will be able to more tightly control network traffic thereby making the network more efficient (forward packets only to their actual destination). The antenna design will consist of a combination Yagi-Quad directional antenna to achieve both the gain and impedance required to efficiently transmit our signal. Range estimates thus far indicate that a wireless link could extend well beyond our two mile target; however more testing is required to get an accurate maximum distance. All of the above will be connected to a properly sized Uninterruptible Power Supply (UPS) to handle both fluctuating input voltage (from the power grid), and complete power loss. The UPS chosen from American Power Conversion (APC) is estimated to provide about twenty minutes of uptime off the shelf, but can be expanded using additional batteries. Our goal is to allow our computer system to remain active for one hour in the event of complete power failure. The diagram below illustrates our proposed overall solution. 13 Project Management Plan Personnel Daniel Newport (webmaster) Kevin Bishop (Presentation/Lab) Steven Sadler (Documentation) Brian Holt (Management) Shared Projects Components Quagi Antenna UPS 2 Computers Filtering Routers Secondary Routers 8 screens 4 Video Cards Operating system Filtering Router software Secondary router Tasks Installing operating system o Set up multi-seat Program installation Programming router o Install Filtering Software Program installation Designing / building antenna Selecting UPS Designing / building antenna Selecting UPS Selecting Routers, video cards Resources Hockey stick Welding rod Nuts and bolts Coaxial cable Connector VR 1500i Think Station S10 Wireless G router Wireless G router Source ECE shop L157 GeForce 8400 GS Ubuntu Linux Open WRT Lenovo NVIDIA Free open source Free open source Open WRT Free open source 14 APS Lenovo Linksys Linksys Costs (Approximate) Computer: Lenovo Thinkstation S10 - $1190 Video Card - $64 Keyboard/Mouse - $30 Lenovo Monitor - $239 Router - $50 Antenna - $70 dollar per if bought Antenna - $20 if built UPS - $550 Software: Ubuntu- $0 Multi-seat Display Manager- $0 Xephyr- $0 X windows- $0 All Software is open source 15 Fast Diagram References Images All images on the cover sheet of this document were obtained from Wikipedia. All other images were created using Microsoft Visio. 16