Fri 04 Sep 2009 14:57

Assessing H1N1 risk
What sort of risk does H1N1 (Swine Flu) present this flu season? To assess this, it might be helpful to estimate some H1N1 risks and then compare it to risks with which we are more familiar.

So let's look at some numbers. The worldwide case fatality rate of H1N1 (the number of people who have died of H1N1, divided by the number of people who have gotten H1N1) has been estimated to be 0.45%. Unlike seasonal flu, roughly 80% of those who have died of H1N1 are less than 65 years old (typically 90% of seasonal flu fatalities are 65 years old or over). If we assume a 15% probability of getting H1N1 this flu season, the likelihood of someone under the age of 65 dying of H1N1 this season is thus 0.15 x 0.0045 x 0.80, i.e 0.054% or 1 in 1852. This is a little less than the one-year general odds of death due to external causes in the US, approximately 1 in 1681.

/misc permanent link

What's Good About Twitter?
Twitter has a mixed reputation. Negative views generally express the notion that Twitter is pretty much useless , or is a massive waste of time. Indeed, there is no shortage of evidence for this view. What is the usefulness of knowing that someone is brushing their teeth, or having cereal for breakfast? Probably not much. The problem is that "What are you doing?", the question that a tweet is allegedly supposed to answer, is often not very interesting. What one is thinking, what one has stumbled across, or what one wants to tell the world, could be much more interesting.

One very useful purpose Twitter serves is to announce new articles, blog entries, events, or news items when they appear. Twitterfeed makes this easy: it will check an Atom or RSS feed periodically, and automatically tweet the titles and URLs of new articles to Twitter, allowing anyone following the tweeter to be made aware of the new material. For example, my department now uses Twitter to disseminate its news and events.

So is Twitter a waste of time? Is Twitter useless? Only if one takes Twitter's "What are you doing?" too literally. Indeed, some seem to feel the need to tell others whenever they're yawning, scratching an itch or drinking coffee. Clearly this is not the most interesting of material. But, on the other hand, if one uses Twitter to follow information sources (people or organizations) with a high information content, and/or to disemminate such information oneself, it can be very useful indeed.

/it permanent link

How well do Java Virtual Machines Scale Up?
Java seems to be a popular language for small to medium-sized applications and its use at that scale is well understood. But what happens when you scale it up to something quite large? It seems that very large Java Virtual Machines (JVMs) are still rather rare. Blackboard is a Java-based learning management system (LMS) now in use at the University of Toronto. The University is rather large, with a great many courses and students, and its Blackboard JVM is correspondingly huge. It turns out that an ordinary off-the-shelf JVM suffered some unusual and unpleasant performance issues (mostly related to garbage collection) when scaled this large. The university and Sun Microsystems eventually resolved the issues quite nicely (using the new and still somewhat experimental Java G1 garbage collector) but it was an eventful journey. John Calvin of the University has put together a rather interesting talk about this, which will be given at the university on June 23rd, and later this summer at BBWorld 2009.

/it permanent link

Understanding Portable Displays
Perhaps the most important thing about a portable computer, be it a notebook, netbook, PDA, smartphone, or internet tablet, is what it provides you versus what it demands from you. One of the most important things a portable machine provides is logical screen area or screen resolution: the amount of data it can show you on the screen at one time. But of the most important things a portable machine requires/demands is weight: what does it take to carry it?

Screen resolution is measured as a grid of pixels (dots) in width x height format, e.g. 1024x768 means a screen that is 1024 dots wide and 768 dots high. Weight is of course not the only thing that determines portability: size is important too, but generally larger machines are heavier and smaller ones are lighter, so weight is a good shorthand for "weight and size".

A quick way to approach the costs and benefits of a portable computer is to compute the ratio of its benefits (e.g. screen resolution) to its portability cost (e.g. weight). So a quick assessment of a portable computer is to compute its pixel to weight ratio: if the weight ratio is high, the portable computer may compare better to one that has a lower pixel to weight ratios. I've written a little tool to compute this information (in units of pixels per milligram, i.e. ppmg), at http://www.cs.toronto.edu/~jdd/screenspec.cgi.

Pixel to weight ratio isn't quite enough, though, because there are limits to human sight: a portable computer is of no use if the pixels are so small that they cannot be easily seen. "Too small" depends on the distance the screen is from one's eyes. I tend to use devices like cellphones and PDAs at a distance of 18 inches from my eyes, and laptops at 24 inches. Generally, distance multiplied by the pixels per inch of the screen is a constant. For example, for me, I'm quite comfortable with 170 ppi at 24 inches, but beyond that, I feel some eyestrain. At 18 inches, that works out to (170 x 24) / 18 = 227 ppi. In my (anecdotal) experience, many people seem comfortable with 125ppi at 24 inches and 167ppi at 18. Of course, there is much more to this than a simple ratio: tolerance for high pixel densities varies depending on what the person is doing, what fonts are being used, and many other things.

Still, a pixel to weight approach lets one compare machines in interesting ways: for example, an Apple Ipod Touch has a 3.5" 480x320 screen and weighs 115g; that's 164 ppi and a pixel to weight ratio of 1.3. This is comparable to a Nokia E90 Communicator, which has a 4" 800x352 screen and a weight of 210g; its ppi is 218 and pixel to weight ratio is 1.34. But now consider a Nokia N810 Internet tablet: its 4.13in 800x480 screen (ppi is 225) and weight of 226g gives a significantly higher pixel to weight ratio of 1.69. But with ppi around 220 vs. the Ipod's 164, either Nokia device may result in eyestrain where the Ipod Touch does not.

Now look at some notebooks. A (large and heavy) Dell Vostro 2510 notebook weighing 5.72lbs with a 15.4" WUXGA (1920x1200) screen offers 147ppi and a pixel to weight ratio of 0.9, which is (perhaps surprisingly) a higher pixel to weight ratio than a (small and light) netbook, the Dell Mini 10 with the new high-resolution 10" 1366x768 screen (ppi of 155); its weight of 2.86lbs results in a lower pixel to weight ratio of 0.8 (at a slightly higher ppi, too). Compare this to a Macbook Air: with a 13.3" 1280x800 screen, it weighs 3 lbs; its pixel to weight ratio is 0.75. Unlike the other too, though, the Macbook air has an easier-to-read ppi of 113.

Of course, this doesn't mean that one should pick portable computers based solely (or even mostly) on pixel to weight ratios, or ppi for that matter. It is merely one possibly useful way to compare portable machines, and should be at most only one criterion among many, when making a decision.

/it permanent link

Why a netbook?
What is a netbook anyway? It's a new type of low-cost ($350-$600) notebook that is particularly small and light: it typically has a 7-10" screen, and a weight not much more than two pounds. Small and light notebooks are not new, but they have for years been quite expensive, marketted to busy executives who want something small and light to use when travelling, and are willing to pay for the privilege. But an inexpensive small and like notebook is new, so new that it has been given its own name: netbook. The rationale behind the name is that this meant to be an "internet" device: its primary role is web browsing, and office productivity applications are secondary. Such a device relies on wireless: wifi for now, but increasingly 3G and other forms of cellular data service.

Why buy one? It's affordable, by notebook standards. It's also very portable: while it's too large for a pocket, it can easily slip into a handbag. And while it may be designed for internet connectivity, it can run modest office productivity applications. But it is limited in various ways: the small screen, while generally bright and visible on most models, does not have a great deal of screen real-estate; typically 1024x600 or less. RAM and hard drive space is generally less than most notebooks or desktops, and RAM in particular is limited to a maximum of 1GB or 1.5GB, depending on the model, enough for Linux or Windows XP, but not generally enough to run Microsoft Vista quickly. It lacks any form of CD/DVD drive. And the internal CPU (generally an Intel Atom or a VIA C3) is slow, and single-core. A low-end laptop can be bought for as little as $500-$600 with a much larger screen, more memory, a built-in DVD-writer and a more powerful CPU. But it will be quite a bit larger and heavier. That in the end is the key question: is the portability of a netbook worth its tradeoffs? Sometimes yes: if one's computing needs are modest but one wants one's computer whereever you go, then portability is paramount. Sometimes no: those with more than modest computing needs will quickly run into the netbooks' limitations. But whether or not a netbook or a notebook is a better fit, it is nice to have the choice, for a reasonable price.

/it permanent link

How to buy a Computer
For years I have been asked for my advice about buying computers. My advice has changed over the years, because computers have changed, but one thing seems to be constant: a great many people seem to be very insecure about buying computers. This leads to a great deal of angst, and sometimes to purchases that are much more expensive than they need to be. But there are a few common-sense principles that are generally constant:

1. Think carefully about how the computer is going to be used.
This is the key principle that overrides all others. A computer is a tool. Tools are useful only when they can be used effectively. Do not choose a computer that does not fit with the way you use computers. For example, if you are a small person and like to work in many different places, a large and heavy laptop, or worse, a desktop, will not be a good choice: it is worth investing in something small, light and easily carried. If you are a gamer, particularly if you plan to invest a great deal of time playing games that require high-performance video, you'd best invest in a desktop with high performance graphics, even if it is expensive. Playing a demanding game on a cheaper machine with poor video performance will be frustrating. But if you merely browse the web and run productivity applications like spreadsheets and word processors, investing in high-performance gaming computing is a waste of your money.

2. If a better option is available for a lot more money, choose it only if you know you need it.
Insecurity about buying computers prompts people to pay a great deal more money for things that they think they might need: particularly a fast CPU (the computer's processing unit) or a high-end computer model instead of a lower-end one. The price difference can be significant: a high-end model can cost 3-4x the price of a lower-end model, and a high-end CPU can more than double the cost again. For example, the base configuration of the lowest-end home desktop with the lowest-end CPU on Dell Canada's web site is currently $329; the highest-end base configuration with the highest-end CPU is $3150. That is an order of magnitude difference in price. Put another way, the high-end configuration is the price of a formal dining-room suite. The low-end configuration is the price of a single chair in that dining-room suite. If you are paying the high-end price, make sure you need what you are paying for.

3. If a better option is available for only a little more money, chose it unless you know you don't need it.
If it only costs $20 to get a little more memory, a bit faster CPU, or a potentially useful device like a webcam, a fax modem, or a media card reader, why not get it, especially if it's much more money and less convenient to get it separately later? An integrated webcam is a $20 option on many laptops; adding later an external webcam of comparable quality that clips onto your laptop may cost you as much as $90. Or, for example, a fax modem may sound like obsolete technology, and it is, but it can be very convenient to send a fax from your computer by simply printing to the "fax" printer and typing a phone number. The one exception here is to watch out for large price increments for tiny increases in hard drive size: the price difference between a 250G and a 320G hard drive should be a on the order of $10, not $60-70. While one may argue that there is perhaps some value in paying a bit extra for the convenience of ensuring that your computer comes with a decently large hard disk, even a small hard disk these days is quite large. Another thing to consider: if the price difference between a notebook and a desktop is fairly small, and there is no compelling reason to choose a desktop over a notebook, just get a notebook.

4. Assess carefully your need for extended warranties.
Extended warranties can be expensive. However, if you are accident-prone (coffee over the keyboard, dropping your laptop), anxious or risk-adverse, an extended warranty may be worthwhile, particularly the sort that covers accidental damage. Note, though, that on average one spends much less over the lifetime of the computer to repair it (often $0) than one would pay for an extended warranty. Such warranties are often bought out of insecurity, and are highly profitable for computer vendors and technology stores. If, however, you do not expect to have free funds to handle an unexpected repair, especially if the computer is particularly expensive, an extended warranty may be worthwhile as a form of insurance.

5. Don't panic. Most of the available options are all reasonable choices.
Most computers are quite acceptable: there are few bad choices. Choosing a computer is most often a matter of choosing the best choice from among good choices. So relax: even if you miss the best choice, you'll probably end up with a perfectly good computer.

6. Don't forget backups.
The most valuable part of your computer is your data. Make sure you have backups of it, so that if something bad happens to your computer, you will not lose your data. You can always replace a computer. Can you replace your data? The easiest way to back up data is to buy an external hard disk and copy your data to it. Buy that external hard disk when you buy your computer. Yes, you can back up to writeable DVDs if you want, or copy to flash memory of some sort, but it can be a lot of work to divide up your data into DVD-sized chunks, and backups that are a lot of work often turn into backups that are not done.

/it permanent link

Why own a Desktop computer?
The thirty-year reign of the desktop computer may be coming to an end. According to various news reports, since about the mid 2000s, notebook (or laptop) computers have been outselling desktops. More surprisingly, perhaps, miniature notebook computers like the Asus EEE PC, with small screens and low-power CPUs that are no more powerful than mainstream CPUs of a half-decade ago are becoming increasingly popular, with a flurry of new low-cost (about $500) models. The reasons are intriguing: few productivity applications such as personal databases, spreadsheets, word processors and presentation tools need more than a small fraction of today's fastest CPUs. Thus the sort of CPU tradeoffs that need to be made to ensure long battery life in a notebook are less and less noticeable in practice. Other tradeoffs are also diminishing in importance: notebook screens can be large and bright, more and more rivalling desktop screens, notebook hard drives can be spacious and increasingly fast, and the rise of USB peripherals has made a portable notebook with a couple of USB ports as easily expandable in practice as any desktop computer. While notebooks are still pricier than desktops, the price difference is steadily diminishing as manufacturing economies of scale begin to weigh in. Even many who are in the habit of using their computer in one spot most of the time are realizing that an external screen, keyboard and mouse can be added to a notebook to make it function as if it were a desktop for general use, but when necessary, the notebook can be used elswhere, providing the convenience of having one's computer along (with all its data and software) when needed, without the fuss of copying data and worrying about different versions of applications. Moreover, notebooks have been improving in those areas where they offer abilities not found in desktops: battery life has steadily increased from the one to two hours common a few years ago to three or four hours. Lightweight notebooks are increasingly available, and not all of them are expensive. Most importantly, various forms of wireless networking are becoming ubiquitous, providing internet connectivity to notebooks without the need for wires. As such, it is no surprise that notebook computers are being widely purchased, and many peoples' first computer is now a notebook, not a desktop.

There are still some good reasons to buy desktops. The lowest-cost computers are still desktops, not notebooks. The very best CPU and graphics hardware is available only for desktops, and many modern games use as much of these resources as they can get. Hence desktops suit hardcore gamers much better than notebooks. Finally, Microsoft Windows Vista generally requires much more CPU and memory than most other operating systems, and the introduction of Vista has put some pressure on computing resources; because of this, some of the less powerful notebooks are now shipping with versions of Linux or with a previous version of Microsoft Windows, such as Windows XP. Nevertheless, it seems clear that given the increasing attractiveness of notebooks in comparison to desktops, a sensible way to approach buying a computer is to simply buy a notebook unless one has some concrete reason to need a desktop.

/it permanent link

Web shell scripting
It is a very handy thing to be able to write a quick script. UNIX and its derivitives have long been superb at making this possible: they possess a great many utilities that are designed to be used both from the command line and within scripts, and they possess shells that have all the control structures one might expect from any programming language. In fact, the traditional UNIX philosophy is to write small programs that do one thing well, and then combine them using scripts into rich and powerful applications. Indeed, the UNIX scripting environment is a rich one. But it is difficult to write shell scripts for the web. The unix scripting environment is designed for files, not web forms, the contents of which are encoded as url-encoded or multipart-encoded data. Hence, while unix shell scripts are sometimes used for web applications (cgi scripts), they are relatively rare, and generally frowned upon. The reason is no surprise: url-encoded and multipart-encoded data is complex to parse, and shell scripts that parse such data using sed, awk, etc. tend to be slow and hard to get right.

But this is easily fixed. If UNIX shell scripts like files, then they should be fed files. Hence I've written a small program (in C and lex), urldecode (ftp:/ftp.cs.toronto.edu/pub/jdd/urldecode) that parses url-encoded and multipart-encoded data, and converts the data into files. No complex file encoding is used. urldecode reads url-encoded or multipart-encoded data, creates a directory, then populates it with files such that each filename is a variable name, and the file contains the variable value. So all a web shell script needs to do to parse url-encoded data is to run urldecode on the data received from a web form, then read the results out of suitably named files. While this is hardly a replacement for PHP or .NET, it does provide a surprisingly simple and straightforward way to script for the web, because it allows all the handy UNIX utilities in the UNIX shell script environment to be leveraged to process web data. That's useful.

/it permanent link

CRT to LCD computer lab replacement: how much of a difference?
We are replacing all the remaining CRTs in our Computer Science teaching labs this summer with LCD panels, a total of 84 units (we replaced about fifty last summer). It's well known that LCD panels generally use less power when displaying than CRTs do, but the question is: roughly how much power/carbon emissions will we be saving through this summer's CRT replacement?

Using a Kill-A-Watt electricity consumption meter, we measured the power consumption of our CRTs (19" Dell P992) and our new LCD panels (19" Samsung 920BM and 22" Samsung 2253BW). When displaying an image, the CRTs draws between 85W and 110W of power, depending on how bright/white the image is. In comparison, the 19" LCD draws 35W of power, and the 22" draws 41W. If we assume an average CRT power draw (when displaying) of 97.5W (the mean), that's a power savings of 62.5W for the 19" LCDs, and 56.5W for the 22". We are replacing 48 CRTs with the 19", and 36 with the 22", for a total power savings of about five kilowatts.

What does this mean over time? If we assume the machines in our labs are displaying for an average of one hour out of eight (our labs are open to students on a seven-day twenty-four hour basis), and if we consider our projected equipment lifetime of four years, this implies about twenty-two thousand kilowatt-hours saved over this period. If we multiply this by a carbon intensity ratio estimate for grid electricity of 0.0453 kgC/kWh, this suggests a savings of about a metric tonne of carbon over the expected lifetime of these LCD panels.

/it permanent link

IT Support and human nature
IT is not about computers, but about people. This may be surprising: after all, when we think about technology, we generally think about equipment, gear, gadgets, code. But this gear doesn't exist for itself alone. Quite frankly, an unused computer is nothing more than a combination of space-heater and white noise generator. The I in IT is information, and that information is generated by, used by, and valued by people. For IT to be effective, it needs to be used effectively. Technology is a tool: powerful and complex, and like all powerful and complex tools, it takes time, effort and a certain amount of talent to learn to use the tool effectively. People are social beings, and so we learn to use tools in a social context: people who "know how" teach and help those who don't. This, broadly speaking, is the logic of IT support, which, ultimately, is a social construct to ensure that those who know how to use IT tools are available to help those who need help to effectively use them.

Human beings live in the tension between the collective and the individual. This is a fancy way of saying that people live by interacting with other people in ways that range from the genuinely interpersonal to impersonal embodiments of complex social constructions. Consider the difference between "I love you" on one extreme and "One Way, Do Not Enter" on the other. Both the collective and the interpersonal elements of human interaction are present in IT support: the nature of the technology imposes the need to interact with complex technical systems, while the nature of the human beings who use the technology requires one-on-one personal interaction. Indeed, IT support fails when it becomes too much like the notion of a "computer centre", too removed from the individual and the person-to-person act of helping and receiving help. But it also fails when it becomes too individualized, because of simple economics: there are many fewer IT experts than there are people in need of their help, and the one-to-one dynamic begins to fail when there are so few on one side of the equation and so many on the other. Effective IT support requires a balance between the two.

One way to maintain this balance, if there is a "critical mass" of IT needs and resources, is to make the commitment to do both at the same time. At the Department of Computer Science at the University of Toronto, we have found an effective way to do this for research computing support. We have a broad and diverse community of researchers, divided up into research groups. They have access to a core IT infrastructure of technical services, equipment and highly skilled staff to run it. But the department also has dedicated IT support staff who partner with specific groups: each group has their own person, their own IT expert, to call upon, and this person knows the people in the group and their research. We call such staff "points of contact", or POCs. Research IT support in the department is not a matter of contacting an anonymous service centre in one's moment of need, in the desperate hope of finding a sympathetic stranger with the requisite skills. Instead, it becomes an interaction between people who know each other, people who have been able to build a trust-relationship over time. Yet the economics of purely individualized support have been overcome: this organization "scales", because POCs do not need to do everything themselves. They and their groups have access to a complete infrastructure that offers common services across the entire department: secure and reliable file storage, web services, email, and more, and the expertise of the skilled technical staff that run it. POCs are freed to focus more fully on the unique, individualized needs of the research groups they serve.

Sounds idyllic, doesn't it? It is, in theory. In practice, there are plenty of challenges. Communication is key: POCs need to communicate well with their groups, and with other POCs and the core infrastructure staff. And the groups themselves need to be responsible for communicating with their POC: in human relationships, even those of IT support, there is both benefit and burden in knowing and understanding the other. A POC who is "shut out" of the research activities of the group is hampered in any effort to provide support that is well-tuned to those activites. That does not mean that poor support will result: even generic IT support, with a human face, can be superior to that offered by an anonymous service centre. But it does mean that the full benefit of having a POC will not be realized. But if a group and a POC fully commit to regular communication, the quality of IT support can be significantly greater than anything from a large service centre, because the POC who is offering that support has the potential to be a creative participant in the group's mission, the very mission that the group's use of IT is intended to serve.

/it permanent link

Blogging: Keeping It Simple
When I decided it was time to start blogging about information technology and information communication issues, I needed to choose some suitable blog software, something that would provide good results but also be easy to use. Open source is preferable. So I decided to do a quick web search and see what I could find. Most blog software seems to be pretty heavyweight: a database on the back-end to hold the blog entries, plus some sort of complex PHP web front end to display them. But this makes no sense to me: why use a database for blog entries? Blog entries are simple bits of text data that are organized in simple ways. There's no need for powerful query languages, transactional locking, and the various good things databases provide. These things are not free: databases have overhead to set up and run - I'm not worried so much about the computational resource overhead, but rather the human overhead: the time and energy required to configure, maintain, and back up databases. Do we really need such a thing for a blog?

Happily, I found blosxom, a piece of open-source blogging software that consists of a simple Perl CGI that uses a simple filesystem backend (a directory hierarchy) to hold all the blog entries. This is a nicely designed piece of software: simple, straightforward, low overhead, and very quick and easy to get going. It's also quite customizable. Clever simplifying details abound: for example, the date of the blog entry is simply the timestamp of the file, you create a new blog entry by simply creating a new file of the right name wherever you wish in the blog directory hierarchy, and writing something into it with your favourite text editor. You can organize your blog entries by topic if you want, and blosxom gets it all right. RSS seems all the rage these days: blosxom does that too: just add /index.rss to the end of the URL. For me, the only annoying bit of this software so far is the spelling of its name: I keep typing "bloxsom" for some reason.

Why is blosxom so good? Because it leverages what is already present on most systems: the filesystem, rather than introduce a complex, powerful and costly tool (a relational database) when it's not really needed. Kudos to Rael Dornfest, who, instead of taking the most popular or obvious approach, took the time to understand the problem being solved and the tools already available to solve it. This is an example of sensible economizing: human time and effort is a valuable commodity, the use of powerful tools (e.g. relational databases) uses up some of that commodity, and so such tools should be avoided unless they are really needed. If you think this sounds a little like "green" or "environmental" thinking, you're quite right: conserving energy to preserve the environment is very similar to conserving human energy to preserve the human environment. Just as the natural environment suffers strains from excessive energy consumption, so the human environment suffers from excessive demands on each person's time and energy. In both realms, economical thinking at design time is a prerequisite to good technology.

/it permanent link