Investigating the Impacts of Web Servers on Web Application Energy Usage (GREENS 2013)
Investigating the Impacts ofWeb Servers on WebApplication Energy UsageComputer and Information Sciences University of Delaware Irene L. Manotas G. Cagri Sahin James Clause Lori Pollock Kristina Winbladh
Which Web Server Should I Use?Empirically Investigate • RQ1—Feasibility: Does the choice of webserver impact the energy consumption of a webapplication? • RQ2—Consistency: Are the web serversconsistent in their impact? 2
Experimental Setupweb browser 3 workloads web server web application LEAP energymonitor Integra+on Tests Automa+c Tes+ng user inputs 3 WEBRick
4 % Difference in energy consumption from the mean Web Servers Feature Mongrel Puma Thin WEBrick Calendar 10.10 -6.10 -8.50 2.30 Context Edit -1.40 -2.10 -0.10 3.40 Preferences -4.00 8.70 -4.00 -1.80 Review -1.10 -6.30 -1.30 7.70 Search 1.80 4.10 5.90 -0.60 Show Statistics 2.70 6.10 -13.90 2.90 Toggle Context -3.00 4.70 7.20 -10.70 Total 1.70 0.10 -3.60 1.70 § A given web server is not always the best under all features. § The web server does make a difference § Energy consumption variability differs across features. 4 4 This work is supported in part by National Science Foundation Grant No. 1216488 and an award from the University of Delaware Research Foundation Results: Feasibility and Consistency
• Correlating energy measurements with designdecisions/implementations in a non-tedious manner 5 Issues We Face Questions for Discussion• How are others monitoring and mapping energy usageto program units? • How many repeated runs do others perform to takemeasurements to account for variations?