The Internet Weather Report forecasts the state of the Internet itself. Lem Bingley finds out how and why MIDS gauges climatic changes on the Internet
The IWR Internet Weather Report is not, as it might sound, a place to learn whether you’re likely to get rained on at the weekend. Unlike the numerous Web sites that attempt to predict the fundamentally unpredictable state of our climate, the IWR reports instead on the state of the Internet itself. Since December 1993 an Austin, Texas-based organization called MIDS – Matrix Information and Directory Services, which is found at http://www.mids.org has been filing a daily report on the state of the Internet’s internal weather. And while the average British weather forecast is concerned with the degree to which the heavens are going to open, the IWR is concerned instead with delays. Specifically, the delays experienced when contacting Internet nodes including servers and routers in different locations around the world. MIDS regularly polls a list of more than 4,000 nodes from its Texas base, using ping. Each ping is sent out five times for each scan, and the entire scan is repeated every four hours. The time spent waiting for each node to respond is measured, this is referred to as its ‘latency’, and the results are then collated and presented graphically on a selection of maps. Each map uses plotted circles to present data in four dimensions: latitude and longitude from the position of a circle; latency from the size of a circle; and finally the number of nodes at a given location with a given latency, gauged from the color of a circle. Larger circles mean longer waits, and so clumps of large circles represent the Internet equivalent of storms – bad weather and sluggish response. Many cities sport concentric circles with colors that show a bell-curve distribution; a few sites with long delays, many with average latencies, and fewer offering rapid response, as you might expect. A further dimension – time – is added by gathering each day’s results together in an MPEG format slide-show. MIDS has been compiling its weather reports more or less continuously since December 1993. It is, therefore, one of the few bodies to have collected any meaningful numbers that might support the contention that the weather on the Internet is getting steadily worse. Some notable observers, including Bob Metcalfe, the inventor of Ethernet, have suggested that the popularity of the Web will lead to a catastrophic failure of the Internet at some not-too-distant juncture, if increases in load continue to outstrip improvements in infrastructure. According to John Quarterman, president of MIDS, the IWR is the wrong place to look to support that argument. What the data indicates is quite the opposite, he states. Not only is the Internet not falling apart, but overall Internet performance has been steadily improving over the entire life of the project, and continues to do so. The near-term forecast is continuing seasonal performance changes, such as higher latencies when college semesters start, and overall continued performance improvements. In fact, the IWR data has convinced Metcalfe to change his prediction from wide- scale collapse to increasing localized disruption. Over the more than three years that the IWR has been providing its intriguing service, Quarterman has lost track of the details of its genesis. It was something that more or less fell out of the other maps we were doing – plus the idea of collecting latencies with ping came from a consulting project, he recalls. The general reason we did the IWR is that we thought it was an interesting idea and we had the capabilities to do it, so we did it. That’s one of the advantages of being an independent research organization. Another reason is that it’s a good thing to give something back to the Internet, which is a fine old Internet custom. Oh yeah: it’s also good marketing, he adds. The reports themselves are funded through sponsorship, and the future development of the service will largely depend on where those funds come from – Italy gets its own detailed map, for instance, because there is an Italian sponsor. But there are also some general improvements on the way. There will be a Java version of the report, which will permit the viewer to select the speed at which the slide-shows play. A later Java version will enable the viewer to see longer animated compilations. Both should show the maps more clearly than the current MPEG format. MPEG is a glossy format that turns out to be better adapted to photographic movies than to data presentations, Quarterman observes. There are plenty of other possibilities, pending funding, he adds. If somebody wants to fund one or more new MIDS IWR features for public viewing, we’re always happy to talk to them about that.