August 9

Contents

Editor's Introduction

This week I report on my conversations with two dozen geospatial professionals about their hardware requirements and point out several New York Times articles about geospatial technology. Plus, 18 press releases.

— Matteo Luccio

Hardware Requirements for GIS

What computer does a typical GIS analyst need? Conversations I had this week with two dozen geospatial professionals confirmed my impression that there are two short answers: "it depends" and "it's a non-issue." Because of the first answer, in what follows I will often use such words as "usually" or "mostly." The second answer is due to the fact that, in general, new software releases are not pushing the limits of hardware bought within the last couple of years.

Hardware requirements for GIS depend primarily on:

• the tasks to be performed
• the size of the files used
• the total volume of GIS work to be performed
• how often the most demanding tasks are performed
• the percentage of time, if any, spent in the field.

Large GIS shops that handle a wide variety of tasks, up to processing very large raster aerial images, need a wide range of machines — from standard PCs to high-end dual processor workstations with half a terabyte of memory. I am focusing here on the needs of typical, individual analysts. For enterprise GIS, the requirements are much more complex, and key variables include the number of users and their average and peek needs, which determine the minimum network bandwidth. (An excellent source on this is Roger Tomlinson's Thinking About GIS: Geographic Information System Planning for Managers, Third Edition, recently published by ESRI Press.)

The reason that, according to several of the people I consulted, hardware requirements for GIS analysts is a "non-issue" is that computers have become so cheap and commoditized that hardware is now rarely the limiting factor. For most GIS professionals, a typical dual-core processor running with a speed of at least 1.8 MHz, at least 1 GB of RAM, and a 100 GB hard drive is sufficient. A 3 GHz processor, however, will speed things up considerably, especially when processing complex vector files and intersecting GIS layers.

For most purposes, the more RAM the better. As one GIS manager put it, "For cartography, you can never have enough RAM." Most of the people with whom I discussed this recommended at least 2 GB and some said 3 GB or 4 GB (though there is some question as to whether Windows XP can really use 4 GB). RAM is even more important than more processor speed, because when a CPU runs out of RAM it writes to the hard drive, which slows it down considerably.

Many vendors specify hardware and network requirements. One good rule of thumb might be to take ESRI's minimum requirements and double or triple them. Always buy more than you need now — processor speed, RAM, hard drive, etc. — because you will need it in six months or a year. If you have a small number of GIS users and a large budget, the easiest thing to do is to simply buy them top of the line machines — whichever has the fastest processor, the most RAM, and so forth.

Hard drives are so cheap that you might as well get one that can store a few hundred GB of data. GIS graphics now have so much detail and flat screen monitors have come down so much in price that it makes sense to buy a 21 inch monitor. Many GIS professionals now also use a utility, such as the one built into Windows XP, that allows them to use two, side-by-side monitors as if they were a single one. Screen resolution, on the other hand, does not seem to be a problem. As one GIS analyst told me, it has not been an issue since ArcView 3.

In some cases, the hardware requirements are driven by a computer's other uses, beyond GIS — such as querying very large databases or crunching large, formula-laden spreadsheets. In most medium and large organization, the IT department decides what computers to buy. Often, it buys dozens or hundreds of computers at a time and it is hard for GIS analysts to convince IT that they need a faster processor and more RAM than others in the organization. (A related issue is whether GIS analysts are on the IT staff or constitute a separate unit. As one county GIS analyst put it to me, "If you are an IT client, you can get the box you want, but if you work for IT you have to fend for yourself.")

A key issue to consider is the extent to which you use aerial or satellite images and how compressed they are. TIFs need a huge amount of disk space, but can be greatly compressed using such utilities as LizardTech's MrSID (Multiresolution Seamless Image Database) or ER Mapper's ECW (enhanced compression wavelet) — a patented technique that displays images quickly and makes good usage of RAM. Using ECW, you can uncompress only the extent you need to see, uncompressing further as you zoom in. geoTIFFs, on the other hand, don't compress images as much and require you to uncompress the entire file. Of course, image compression is an inherently "lossy" process; however, depending on what you are trying to do, you can often compress an image at a 20:1 ratio and hardly notice the difference. If you are dealing with an image at 1:10,000 scale, you can probably compress it as much as 40:1. (Grayscale images can be compressed less.)

Some really large jobs require extra processing power. One analyst recently used LizardTech GeoExpress 6.0 to mosaic 36 quads into a single image to print a large map. With 1.5 GB of RAM and 15 GB of processing space, the task took him 1,200 hours. If the system had crashed, he would have lost weeks of processing time. He argues that standard-issue desktop computers are "paper weights for secretaries and managers" and that serious GIS analysts really need workstations with a direct connection to a printer, rather than having to go through the network. The reason that IT departments usually resist this arrangement, he says, is not so much the cost as the fact that it does not allow them to handle maintenance and upgrades from a single location but requires them to make "house calls" to individual work stations.

The manager of the GIS shop for a metropolitan area was trying to intersect tax lots with a zoning layer on a single processor, but it kept crashing and would never finish. So he put the job on a server-scale machine, with a quad-core processor and 4GB of RAM, and it ran in 45 minutes.

For a GIS shop that specializes in aerial imagery, the tasks that take the longest are ingesting the imagery, triangulation, and orthorectification. The whole process has seen tremendous growth in the past five years, partly because of the increase in the availability of satellite imagery. One production person told me that he used to triangulate only 20 or 50 images a month from a single sensor, but now he is handling thousands from a variety of sensors.

Often GIS analysts, in addition to GIS software, also need to use a browser, a spreadsheet, and other programs. Dual processors have greatly relieved this challenge: as one GIS analyst told me, even when his GIS program is using one processor to capacity, he can still use the other to check his e-mail. However, ArcGIS only takes advantage of one CPU at a time.

The amount of RAM you can use skyrockets if you switch to a 64-bit operating system. However, a lot of software does not yet work reliably on such systems. One IT manager told me that his organization uses 64-bit servers running with ArcSDE but has to use work-arounds. Another one told me that, for one analyst, he had to install a 32 bit system on top of a 64-bit one, because an ArcMap extension did not work on the 64-bit system.

As GIS goes increasingly mobile, the key issues in the field are portability — weight and size for a given amount of battery time; connectivity — to other devices, via Bluetooth or cable, and to wireless LANs; and compatibility with GPS receivers.

As for portability, for several years the trend was to integrate data collection, storage, and display into a single hand-held GPS receiver. However, a more modular approach, while forcing you to carry more devices — a GPS receiver, a notebook computer or PDA, a cell phone, and flash memory cards or external hard drives — allows you to stay more current, by upgrading them one at a time, as needed. A rugged laptop for field use has a life cycle of a couple of years, while a GPS receiver can last twice as long.

As for connectivity, now even high-end PDAs have WiFi. However, a high speed wireless connection allows you to transfer data in real time and to access, for example, an image server. Good connectivity also compensates for one of the biggest limitations of field GIS: the amount of data that a handheld device can store. Of course, your connection from the field over a cell phone network is still much slower than via an Ethernet connection to your office LAN. Still, you can now display Google Earth imagery behind your map on a PDA in the field!

As for compatibility, most off-the-shelf data capture software will support the NMEA data stream from different GPS receivers. While most GPS receivers still require a cable connection, several now offer a Bluetooth option. The modular approach allows you to use a consumer-grade, resource-grade, or sub-meter GPS receiver, as needed. All pocket and tablet computers now come with Bluetooth, though you may have to specifically ask for that option on tablet computers.

Few GIS analysts, it seems, use laptops, except for fieldwork, presentations, and demonstrations. As one GIS analyst put it, "laptops are for doing your homework in a coffee shop."

What's the next frontier in hardware for GIS? Several of the people I consulted had the same answer: now that multi-core CPUs are the standard, the next step is the ability to use them intelligently, for parallel processing.

Briefly Noted

I use coverage of geospatial technologies in the New York Times as a rough indicator of the extent to which they have reached a mass audience. In the last few weeks, I have come across several articles on mapping, GPS, and other geospatial technologies and issues.

July 5 — "Traffic Alerts Get Personal, With Made-to-Order Data," by Jacques Steinberg, describes various services, including Navteq's Traffic.com. "In the latest incarnation of traffic reporting," Steinberg writes, "information gleaned from strategically placed cameras, road-top sensors, electronic tollbooths and eyewitnesses is edited in Mission Control-style command rooms, and sent via personalized text or voice messages to subscribers' cellphones or BlackBerrys, often at no charge."

July 5 — "Lost? A Personal Locator Beacon Could Save Your Life: Calling for a Rescue At the Touch of a Button," by Dan Mitchell. "When activated," he writes, "the personal locator beacon emits a 406-MHz radio signal that is picked up by one of 12 satellites operated by the international Cospas-Sarsat search-and-rescue system. The signal carries a code that identifies its owner and, depending on the completeness of the required registration, the owner's emergency contact information and expected location. Older or less-expensive personal locator beacons use Doppler radar to determine the user's location. Newer, costlier models include Global Positioning System technology, which pinpoints the user's location faster and more accurately." The article reviews the history of the devices, what happens when they are activated, and some of the models available.

July 16 — "When a Phone's Trill Prompts the Clang of Prison Doors," by Anemona Hartocollis, describes the growing use of cell phone records in criminal trials, by both prosecutors and defense attorneys, to prove the location of a defendant at the time of a crime.

July 19 — "Global Positioning by Cellphone: New Services Track Children, Locate Concerts and Blaze New Trails," by Larry Magid, describes a dozen products and services.

July 27 — front page article, "With Simple New Tools on Web, Amateurs Reshape Mapmaking," by Miguel Helft. It tells a story very familiar to readers of GIS Monitor but not to the general public: "With the help of simple tools introduced by Internet companies recently," Helft writes, "millions of people are trying their hand at cartography, drawing on digital maps and annotating them with text, images, sound, and video. In the process, they are reshaping the world of mapmaking and collectively creating a new kind of atlas that is likely to be both richer and messier than any other. They are also turning the Web into a medium where maps will play a more central role in how information is organized and found." The article describes several examples of this new popular mapmaking and compares these maps, in aggregate, to Wikipedia. It describes the GeoWeb as partly a byproduct of the competition between Internet search companies and their decision to allow programmers to create mash-ups. "For the Internet search companies, these efforts are part of a race to capture the expected advertising bonanza that will come as users browse through these maps on their computers or cell phones in search of businesses and services. In the process, they are creating technologies whose impact could be similar to those of desktop publishing software, which turned millions of computer users into publishers."

August 5 — "Think Your Dog is Smart? Its Collar May Be Even Smarter," by Anne Eisenberg. "Many dogs wear collars with ID tags. Now some collars," she writes, "also have Global Positioning System units, motion sensors or other additions to help owners keep track of their pets the high-tech way. Garmin, a manufacturer of G.P.S. equipment, makes a tracking system that keeps tabs on dogs during walks in the countryside or in the dense ground cover of a hunting trip. It has two parts: a hand-held G.P.S. unit for the owner and another device that is mounted on the dog's collar or harness."

August 7 — obituary for "Jack Cole, 87, Creator of People Locator," by Margalit Fox. "Jack Cole," she writes, "[...] used early computer technology to sort the world — or at least millions of people in it — by street address, creating a series of reverse directories that remain invaluable to detectives, debt collectors, telemarketers, and anyone who needs to find someone[.] [...] Sixty years ago, Mr. Cole began publishing the Cole Directory, a set of reverse guides to various U.S. cities. Known familiarly as crisscross directories, Cole Directories list a city's residents by address and by telephone number. The first city directories in the United States were published in the 1780s. Though most were arranged alphabetically by last name, a few early ones were organized by street address. For the next century and a half, compilers of these directories trudged door to door, painstakingly recording the residents of every apartment in every building on every block in the city. What Mr. Cole did, starting in 1947, was to use I.B.M. punch cards to streamline the process, turning an ordinary telephone book into what today would be called a searchable database."

» Back to our GIS Monitor AUG 2007 Issue