Managing a Wi-Fi network is like bringing cake to the office, wrote Aditi Bartl, director of strategic communications for Stanford University IT (UIT). “The more people who show up for cake, the smaller the slice each person gets.”
Stanford installed its original Wi-Fi network at Main Quad in 2001. As more people ventured into cyberspace, the capabilities of the network needed to expand in return. By 2004, Stanford’s campus had roughly 545 access points and more than 1,100 concurrent users at any given point during the daytime. Today, some 33,000 devices access Stanford’s network through one of 17,000 Wi-Fi access points per day.
Stanford Wi-Fi tends to perform at par or above the fundamental recommendations of the U.S. Federal Communications Commission based on an observed mean download speed of 100.43 megabits per second in an analysis by The Daily.
The observed median download speed of 69.11 megabits per second is considerably lower and may agree more closely with students’ day-to-day experiences using the network.
To assess the current performance of the Wi-Fi network at Stanford, The Daily collected 65 internet connection samples — each of which included download speed, upload speed, latency and packet loss — using the Stanford Wi-Fi network across campus from Sept. 28 to Oct. 12, observing significant variations in performance by location.
This survey does not include samples from the Gates Computer Science Building, in which the Department of Computer Science installed its own Wi-Fi separate from UIT.
According to Bartl, the separate Wi-Fi is due to the faculty’s networking research. It is the only department with its own Wi-Fi network.
The mean and median download speeds across all sampled locations were 100.43 and 69.11 megabits per second, respectively, with a peak speed of 359.68 megabits per second, observed between Encina Hall and Knight Management Center.
The slowest download speed of 0.6 megabits per second was recorded at an outdoor bench between the Gates Computer Science Building and Campus Drive, and the second-slowest was 6.06 megabits per second just south of the Gunn Building. Both of these samples were taken outside rather than inside a building.
The FCC’s Broadband Speed Guide provides the minimum speed required to stream 4K video as 25 megabits per second. Lightweight web browsing and email activity can be accomplished with a speed of one megabit per second, according to the guide.
The FCC considers an Internet connection high-speed, also termed “broadband,” if it achieves download speeds of at least 25 megabits per second and upload speeds of at least three megabits per second, though those definitions are from 2015 and may change at a future date.
The distribution of speeds revealed that 80% of the campus locations achieved speeds above the FCC’s broadband requirement of 25 megabits per second, indicating a robust network performance in those areas. More than 40% of samples exhibited speeds above 100 megabits per second.
On the other hand, the latency and packet loss metrics provided further insights into the network’s responsiveness and reliability — the lower the latency, the better the responsiveness.
Latency is the delay between a user action and the network response. The average latency across the campus was 5.15 milliseconds, with the highest latency of 14 milliseconds observed in the Jerry residence by Lake Lagunita.
In computer networking, when digital content is sent through a medium, it is broken down into small units, or “packets,” of information. Packet loss represents the percentage of these units that don’t successfully make it to the end of the journey from transmitter to receiver. A critical metric for real-time applications like video conferencing, it averaged to 0.64% in the samples.
A comparative analysis between download and upload speeds revealed a correlation coefficient of about 0.87, indicating a strong relationship between the two metrics.
The infrastructure for Stanford Wi-Fi is supported by three teams in UIT: LAN Engineering, Installation & Maintenance and Field Engineering. Among these teams, the most common challenge in everyday work is that “[w]here Wi-Fi equipment needs to be placed to be effective isn’t always pleasing to the eye,” Bartl wrote.
According to Bartl, installing equipment in student residences can be particularly difficult.
Chris Davalos ’24 spent winter quarter of his frosh year in Roble Hall where he relied on a “really dodgy” Wi-Fi connection to keep up with his coursework. Amid COVID-era restrictions where remote learning was the dominant mode of instruction, Davalos said his connection would sporadically drop off, sometimes for hours at a time — as a result, he missed some classes.
Davalos has since lived in two different Row houses, where he has noticed a general improvement in Wi-Fi performance. He said that it would be helpful to know more about the services that the UIT Help Desk offers and where students can provide feedback.
“Usually you just toughen up,” Davalos said in reference to everyday technical difficulties in connecting to the Internet. “Is that something we should be reporting so [UIT] know[s], or do they already know?”
A wired connection will always perform better than a wireless one, Bartl wrote. With a wired connection, such as that through an Ethernet port in a dorm, there is one dedicated connection for that device, and its strength is not affected by whoever else is online.
For wireless connections, getting online can be a zero-sum game for individuals who are remotely (and perhaps unknowingly) sharing the same medium — the Wi-Fi access point — depending on the kinds of software that are accessing the network.
Last year, Christian Gebhardt ’24 and his roommate set up their own wireless network after two weeks of being unable to connect to Wi-Fi in EVGR-A Duan Family Hall.
“[We] filed a report saying we didn’t get Wi-Fi — issue never changed,” Gebhardt said. “There wasn’t anything [UIT] could do about it.”
Though they could have plugged their devices directly into the local area network without a router, Gebhardt said that he only found ports in the common room of his EVGR apartment, rather than the bedroom.
Gebhardt also noticed a weak cell connection in EVGR-A, which prevented him from taking calls.
“If you’re walking through the dorm, you’ll lose cell service,” he said. “So that was annoying … you could kind of mitigate it if you used Wi-Fi calling, but our room didn’t have Wi-Fi.”
According to Bartl, older Wi-Fi equipment tends to fluctuate in performance due to the nature of their design, but UIT is upgrading the Wi-Fi in several locations, especially densely populated areas.
“We’re currently upgrading the campus to Wi-Fi 5 or better, and are working to simplify, enhance and expand Wi-Fi access and availability for students, faculty and staff,” she wrote.
The upgrades are expected to enable newer devices to utilize the better protocols of the day, including Wi-Fi 6/6E. The Wi-Fi 6E wireless standard comes with the latest connectivity optimizations and can operate on 2.4-, 5- and 6-gigahertz radio bands.
For official resources and information on Stanford Wi-Fi, Bartl pointed to the Stanford Wireless Network and Services web page, which includes links to the help ticket form, ongoing projects and dedicated status maps for indoor and outdoor network coverage.