WiFi Speeds of Odegaard Undergraduate Library

[INDE 316] 

Ana Lai 

Eric Kammers 

Zach Burgess 

Angela Funk

Abstract 

Odegaard Undergraduate Library at the University of Washington experiences a large fluctuation of foot traffic during the week and must accommodate influxes of students who utilize this shared space to study. Because many students use technological devices that require a reliable and effective internet connection to perform daily tasks, the main concern of this study is to understand if there is an optimal time to be at Odegaard to achieve the best internet connection speed. The metrics utilized for internet speed in this experiment are upload speed and download speed. To collect these measurements, project team members all downloaded the application Speedtest, made by Ookla, onto their respective smartphones. With Speedtest, each member (at different designated times of the day) collected data through the application on their respective phone’s at the entrance of the library. The results of this experiment suggest that the time of day plays a significant role in upload and download speeds of Wi-Fi connection in Odegaard with optimal time interval being 5-10 PM.

Introduction

Odegaard Undergraduate Library (OUGL) is a popular study spot for many students across campus with “a wide range of services and spaces to support undergraduate learning and research” [1]. The library, like other buildings on campus, is equipped with Wi-Fi accessible to all UW students, faculty, and staff. Students say they are able to download and play League of Legends without a problem due to the fast internet speeds. Most students in Odegaard utilize the Wi-Fi through various devices, and a fast connection is important within a large, crowded library to complete various tasks.

Open 24 hours per day, Sunday through Thursday, Odegaard Library is one of the busiest research and study locations on UW’s Seattle campus experiencing over 10,000 entrances per day [3]. As of Autumn 2014, UW’s Seattle campus enrollment achieved 44,786 students, with undergraduates making up 65.7% (29,468) of the total [4]. As enrollment to the university is only expected to increase, Odegaard needs to meet the demand of the influx of students by providing reliable and effective internet connection and speeds to the many visitors who come to the library daily.

The mobile application used to collect the information for the experiment, Speedtest, is developed by Ookla. Ookla is a Seattle based company that gathers speed test data across the world and uses this data to find a better picture of how internet speeds vary across the world. Additionally, Ookla helps provide transparency to internet customers by presenting on their website average speed test results of all internet service providers, based on location [5].

The primary objective of this experiment is to find the optimal time to be at Odegaard to use the internet services provided. The null hypothesis states that there is no significant effect of the time of day on the upload and download speeds. Thus, the alternate hypothesis states there is a significant effect of the time of day on the upload and download speeds.

Methodology

Experimental Unit:

The observational unit used in this experiment is the number of megabits per second, measuring how many million of bits are transmitted or received every second. Note that this is commonly shortened to Mbps, and is not to be confused with MBps, megabytes per second, as a byte is composed of 8 bits.

The dependent variables in this experiment are the upload and download speeds measured by the Speedtest app. The independent variables are the time of day and day of the week the data was collected.

Experimental Design:

4x5 Two-Factor Factorial Design:

• Factor A: Time of Day (7:00 - 10:00 AM, 10:00 - 2:00 PM, 2:00 - 5:00 PM, 5:00 - 10:00PM are indexed at 1,2,3,4,5 respectively)
• Factor B: Day of Week (Monday, Tuesday, Wednesday, Thursday, Friday are indexed at 1,2,3,4 respectively)

• Statistical Model for Download Speed is:

• Statistical Model for Upload Speed is:

Equipment or Software:

Data was collected from three iPhone users and one Sony Xperia user, running Android. Due to group limitations, the effect of using different hardware is unable to be investigated. All four smartphones were equipped with Ookla’s Speedtest mobile application available through the app store of each phone.

Procedure:

Data was collected for five days, starting on Monday, February 22nd through Friday, February 26th. Saturday and Sunday were excluded due to limited building hours and because the number of people in the building would likely be much less than during the school week. UW is also comprised of many students who commute, so we assumed by excluding data from the weekend, times when commuters would not be on campus, we would have more accurate upload and download speeds.

To conduct the experiment, group members had to launch the Speedtest app on their phone, and collect 5 observations consecutively, on the main floor of the library’s entrance. The app is simple to operate, users have to push a single button and the app will run a download speed test followed by an upload speed test and save the results to the app. Group data was recorded using Google Sheets and exported to a CSV file to complete analysis using R.

Results

It was first tested to see if the download speed in OUGL meets or exceeds the FCC’s definition of broadband internet, which was increased to 25 Mbps from 4 Mbps in 2015 [2]. Using a one-sided t-test, the null hypothesis that the mean download speed equals 25 Mbps is rejected and the alternative hypothesis that the download speed is greater than 25 Mbps is accepted. This concludes that Odegaard Library exceeds the FCC’s definition of at least 25 Mbps broadband internet.

Using a t-test and an F-test, it was found that the mean upload and download speeds are not equal and that variances are not equal.

By performing an analysis of variance test separately on the upload and download speeds, factors that affect the speed can be identified.

Analysis of Variance for Download Speed:

Fisher’s Least Significant Difference Test: 

Boxplots show that that the download speed in the evening is much faster than other times of the day and that the download speed is consistent through the week.

Analysis of Variance for Upload Speed:

Fisher’s Least Significant Difference Test:

Boxplots show that upload speed in the evening is significantly higher than earlier in the day and there are small differences in the upload speed through the week.

Discussion

From our results we can conclude that the optimal time to be in Odegaard Library is from 5 PM to 10 PM. From our data, it is evident that time has a significant effect on the download and upload speeds, while the days of the week did not. Days of the week showed significance at a p-value of .106 for download speed and .323 for upload speed. Time interval showed significance at a p-value of 2e-16 for both upload and download speed. While Odegaard has proven from our data that the WiFi connection exceeds FCC standard, we also see that upload and download speeds are not equal and have very different variances.

In the beginning of our experiment, we expected that time would not have a significant effect on the download and upload speeds; however, after looking at the proportion of people who live on and off campus, our data begins to make more sense. 76% of UW students live off campus [4]. Most of these commuters try to beat rush hour by coming to school in the early morning and leaving campus before 5 PM. This may have attributed to our data having means within 10 Mbps from early morning to the late afternoon while spiking up almost double the download and upload speed throughout the late evenings, when most commuters have gone home. For the 11,000 students who do live on campus, including not only the people living in the dorms but also in greek affiliates, university-owned apartments and other housing options, they can expect WiFi connections of up to three times the average download and upload speeds in most homes in America.

While the download and upload speed differed greatly during the different times of the day, our data shows there was small variation between each weekday. With this, we can accept our null hypothesis and expect similar connection speeds no matter the day of the week. We also believe that the difference in the upload and download speeds can be contributed to UW students downloading more content than uploading while using the WiFi connection.

There were many possible sources of error when collecting data for our experiment. The majority of communication between group members was through a facebook group chat. This could have led to different members interpreting the experimental procedure slightly differently. Though we agreed to collecting data at the entrance of Odegaard library, some members may have been at different places near the front door entrance, possibly affecting our data. We were also all restrained with time, since many people in the group are taking hard course loads on top of their internships or part-time jobs. Depending on the person and their schedule, the time intervals varied from three hours to five hours. A five-hour interval may have huge fluctuations in data compared to those of a three-hour interval. Another source of error is the operating system and device that varied between each group member measuring the different speeds.

To have the most accurate results, many changes can be made to this experiment. If we had more resources available, having a robot or computer program that consistently takes speedtests throughout the day will give us a more accurate reading than taking one speedtest for a five hour interval time. We could also track the amount of people in the library to see if the amount of people also affect the download and upload speeds like we expect from the vast UW commuter population. By having a specific location of the speed test and conducting it at the exact same time every day instead of within the interval times, we can also gain more accurate results. It would also be interesting to track the WiFi connections over the length of the quarter to see the significance of different weeks during midterm and finals on the download and upload speeds.

Conclusion

        Through our experiment, it was concluded that the Wi-Fi speeds of Odegaard Undergraduate Library, upload and download, are not equal at various times of the day. Data was collected for five days, starting on Monday, February 22nd through Friday, February 26th with three iPhones and one Sony Xperia. Each group members had to launch the Speedtest app on their phone, and collect 5 observations consecutively, on the main floor of the library’s entrance. Using a t-test and an F-test, it was found that the mean upload and download speeds are not equal and that the variances are not equal. Also, by performing an analysis of variance test separately on the upload and download speeds, factors that affect the speed were identified. The primary objective of this experiment is to strive to find the optimal time to be at Odegaard to use the internet services provided, and we have come to the conclusion that the time period of 5:00PM-10:00PM on weekdays is optimal.

Appendix 

Data files and R code are attached in separate files.

Bibliography

[1] University of Washington. “Odegaard Undergraduate Library & Learning Commons”, UW. Available: http://www.lib.washington.edu/ougl/ [Accessed: 01- Mar- 2016].

[2] M. Singleton, "The FCC has changed the definition of broadband", The Verge, 2015. [Online]. Available: http://www.theverge.com/2015/1/29/7932653/fcc-changed-definition-broadband-25mbps. [Accessed: 05- Mar- 2016].

[3] R. Stacey, J. McKinstry, J. DeCosmo, G. Gallardo, D. Notkin, K. Schlessman, P. Schreiber, J. Webster, B. Wakimoto, K. Woody, H. Gillanders. (2010, March). “Odegaard Undergraduate Library Building Vision Steering Group: Report to the University of Washington Office of the Provost”. University Of Washington.         Retrieved from: http://staffweb.lib.washington.edu/units/ougl/odegaard-undergraduate-library-building-vision-steering-group-report/at_download/file. [Accessed: 06- Mar- 2016].

 [4] University of Washington Office of Admissions. (2014, Autumn). “Quick Facts”. University of Washington. Retrieved from: https://admit.washington.edu/QuickFacts#enrollment. [Accessed: 06- Mar- 2016].

[5] Ookla. “Ookla Speedtest: Because Speed Matters”. Speedtest.net. Retrieved from: http://www.speedtest.net/about. [Accessed: 06- Mar- 2016].