Scott Fazackerley

Cluster head selection using RF signal strength

The LEACH algorithm for selecting cluster heads is a probabilistic method which produces clusters with a large variation of link distances and uneven energy consumption during the data transmission phase. To address this issue, a RF signal strength algorithm based on link quality is presented. Using a competitive distributed algorithm, nodes attempt to reduce the overall energy required for transmission in addition to forming favourable clusters based on Received Signal Strength Indication (RSSI) density and quality. Cluster heads form in areas of high node density leading to a significant reduction in transmission link length, a reduced variance in link length distribution and greater opportunity for energy savings through data aggregation. Simulations show that cluster heads selected by this algorithm form clusters with a lower average link length and have less link distance variability. This produces a lower and more evenly distributed energy cost per node in the network.

Reducing turfgrass water consumption using sensor nodes and an adaptive irrigation controller

This paper describes a complete wireless sensor and irrigation control system that reduces water consumption for residential turfgrass irrigation. It has been estimated that 50–75% of residential water use is for irrigation. Current systems are exceptionally poor at adapting irrigation to meet demand, primarily due to incomplete information for system operators who rely either on visual inspection or periodic irrigation programs. This results in over-watering and fertilizer and soil leaching. Our approach couples easy-to-deploy wireless soil moisture sensors nodes with an adaptive irrigation controller that waters on demand without user input. The result is a system that requires less user intervention, lowers water consumption, and adapts to changing climatic conditions.

Time series compression for adaptive chart generation

Web content is consumed on smart phones, tablets, and computers with a significant variation in device display resolution. Visualizing data is typically performed by extracting data from a database, packaging it as JSON or XML for transmission to the client and then visualizing with a client-side JavaScript library. A major challenge is to retrieve only the required data for visualization. Current approaches require programmers to manually modify their data extraction queries and do not adapt to client display characteristics. The contribution in this work is a configurable data compression method that automatically adapts the amount of data transmitted for client-side visualization based on device characteristics. We evaluate several different techniques for time series summarization and compression and show that the amount of data transmitted can be reduced by between 40% and 80% while preserving pixel-perfect visualization. Reducing data transmitted improves client responsiveness and allows flexible and responsive web content without programmer intervention. The approach is tested on a wide variety of data sets and is implemented as an add-on for a JavaScript visualization library.

Key-value store implementations for Arduino microcontrollers

The Internet of Things relies on connecting devices big and small to the Internet and facilitates their communication and management. This includes the smallest microcontrollers and embedded processors that perform simple but critical tasks. There are many applications where there is benefit to processing data locally on the device before sending it over the network. Since network communication is an order of magnitude more costly than local data processing, efficient techniques for storing, searching, and filtering data are useful. In this work, we present a library of implementations of key-value stores for use on Arduino devices. The goal is to make it easy for Arduino programmers to manipulate data without worrying about implementing data structures and query libraries. Key-value stores are conceptually simple for programmers to use. This paper describes the implementations and provides insights on their performance and trade-offs. The work has being released as open source to the Arduino community and is available at https://github.com/iondbproject/iondb.

GPU accelerated AES-CBC for database applications

Data warehouses and database applications use encryption to protect and secure sensitive data. Encryption and decryption of data during query processing can incur a significant system overhead and reduces the overall throughput of the system as the CPU must perform the operations. Graphic processing units (GPUs) have been shown to improve query processing and compression by executing some of the operations instead of the CPU. In this work, we show that it is possible to use a low-cost GPU as a cryptographic co-processor that significantly reduces the time required while freeing the CPU for other tasks. Database applications process data in pages, and page-level encryption/decryption is fully parallelisable if the pages are independently encrypted. Our GPU implementation is between 3 and 14 times faster than a commonly used CPU implementation and has encryption and decryption bandwidth over 3 times higher.

A flash resident file system for embedded sensor networks

Many embedded devices, especially those designed for environmental sensor logging, have extremely limited RAM, often less than several kilobytes. Logged data is stored on flash memory and needs to be easily managed at low energy cost. A file system is required to efficiently manage the device, specifically dealing with wear leveling of the flash memory. Previous flash file systems, even those designed for small memory devices, still consume a reasonable amount of RAM (1K or more). In this paper, we present a flash file system that supports record level consistency with the entire file system and address mapping functionality stored on flash memory. Although this results in a marginally higher read cost, RAM utilization is less than 150 bytes and the read cost in terms of energy usage is less. The key idea is that NOR flash used on these devices supports direct byte reads not supported by NAND memory which allows page translation and data storage to require less memory and consume less energy.

A holistic framework for water sustainability and eduction in municipal green spaces

This paper describes an irrigation control system that incorporates wireless soil moisture sensors and real time user feedback. The system is deployed on a large scale in a municipal green space. It is estimated that between 60% to 70% of municipal water is used for irrigation. Current systems struggle to adapt to site specific variations. Our approach augments an existing commercial water budget controller with a wireless soil sensor network and an adaptive irrigation controller to significantly reduce water usage. The system requires less user intervention than todays commercial systems that attempt to offer similar water savings. The community is able to watch dynamic changes in soil moisture and watering so that they can garner a better understanding of patterns of sustainable water use.

Write improvement strategies for serial NOR dataflash memory

Embedded systems are ubiquitous and perform tasks such as data logging and monitoring. For these devices, lifetime, power use, and data consistency are critical. Systems require robust and energy efficient storage strategies. Serial NOR Dataflash is commonly used, but suffers from high write and erase times as well as limited lifetime. This work proposes write strategies for serial NOR Dataflash that improves efficiency and power use, and decreases write times. Experimental results demonstrate that using masked overwriting strategies can improve write times by an order of magnitude and reduce the number of required page erases, reduce energy consumed by writing and reduce data transfers by up to 90% for specific applications.

TEFS: A flash file system for use on memory constrained devices

A file system is used to manage data on storage media. The FAT (File Allocation Table) file system was originally designed for floppy drives that were less than 500KB in size, and these drives were not capable of fast random reads and writes. FAT has been adapted to work on other types of storage devices since, and it is still widely used today. It is the standard file system used by microprocessors and embedded devices with constrained resources. Micro-controllers, like the Arduino, only officially support the FAT file system when interacting with a SD card. FAT performs well when data is read or written sequentially, but when data is read or written randomly, there is an impact on performance for large files on page based flash devices that cannot utilize caching strategies. Applications that perform random reading and writing are impacted by this architectural issue. For example, flash data structures, like a B-tree, will have poor performance since random reading is utilized to look up values. TEFS (Tiny Embedded File System) uses a simplified tree indexing structure to take advantage of the fast random reads and writes of flash storage and guarantees that the number of page reads and writes will stay constant as the file size increases when randomly reading or writing. Experimental results show that TEFS has significantly better performance than FAT on the Arduino for random I/Os, and the more efficient TEFS page interface is even slightly faster than FAT for sequential reading and writing.

Reducing data transfer for charts on adaptive web sites

Web content is consumed on devices with a significant variation in display resolution. Visualizing data is typically performed by extracting data from a database for transmission to the client and then visualizing it with a client-side Javascript library. A major challenge is to retrieve only the required data for visualization. Current approaches require programmers to manually modify their data extraction queries and do not adapt to client display characteristics. The contribution in this work is a configurable data compression method that automatically adapts the amount of data transmitted for client-side visualization based on device characteristics. We evaluate several different techniques for time series summarization and show the amount of data transmitted can be reduced by between 40% and 80% on standard data sets while preserving pixel-perfect visualization.