Analyzing the Impact of Delay and Packet Loss on Google Docs

Software as a Service allows end users to use complex soft- ware directly from their browsers, transferring heavy computation to servers in the cloud. One use of this paradigm is word processing, former a classic use cases of Thin-Client computing. Similar to Thin-Client sys- tems, the network parameters are an important influence factor for the cloud application performance.
In this paper, we study Google Docs as an example for online word processing tools. We consider a traditional single user scenario as well as a collaborative scenario with two users working on one document si- multaneously. We identify multiple relevant sub-processes per scenario as performance metrics and use a testbed to automatically evaluate the per- formance of Google Docs under varying network parameters. The main contributions of the paper are (1) the quantification of the impact of network parameters such as delay and packet loss on application per- formance metrics for both scenarios and (2) linear regression models to derive the application performance for giving network parameters.

1. Analyzing the Impact of Delay and Packet Loss on Google Docs Lam Dinh-Xuan , Christian Schwartz, Matthias Hirth , Florian Wamser , Huong Truong Thu

2. Motivation Cloud office application is a possible replacement of PC-based office Managing the quality of network is necessary Retransmit Packet loss Delay

3. Google Docs – Usage Scenarios In single user scenario, one user edits a document individually In the collaborative scenario, more than one user edit a document

4. Emulating the User Interactions Selenium web-driver is used to generate user interactions Selenium is a suite of tools to automate web browsers  A sub-process is considered as an individual user interaction Login Creating Typing Saving Logout

5. Collaborative Scenario Workflow Login Login User 1 User 2 Saving Finish Creating Open Link Share link Typing Receiving Open editor Logout Logout

6. Objective To analyze the influence of network parameters on the performance of Google Docs in two usage scenarios To evaluate the impact of the combination of different network parameters on the behavior of Google Docs Retransmit Packet loss Delay

7. Testbed Setup A dedicated testbed is used to analyze the impact of different network parameters on the performance of Google Docs The network emulator adapts different network parameters to analyze the behavior of Google Docs Window-based vSphere Client Google Docs Control PC VM1 ESXi VMWare Server Internet Network Emulator SSH Protocol Measure: Processing Time

8. Window-based vSphere Client Google Docs Control PC VM1 ESXi VMWare Server VM2 Internet Network Emulator SSH Protocol Measure: Processing Time Testbed Setup A dedicated testbed is used to analyze the impact of different network parameters on the performance of Google Docs The network emulator adapts different network parameters to analyze the behavior of Google Docs

9. Measure the Duration of P rocesses Login Creating Typing Saving Logout           The duration of a sub-process is calculated by the difference of its ending and starting time Login Creating Typing Saving Logout Login Open Link Receiving Logout User 1 User 2                  

10. Single User Scenario – Measurement of Short Text Login and Creating sub-processes are the most delay sensitive. It is due to the Login and Creating rely on multiple serial communications between client and server Typing sub-process is kept almost stable. It is due to the updates to the server are sent asynchronously Saving and Logout rely on few communications with the server

11. Collaborative Scenario - Typing and Receiving The duration of Typing and Receiving sub-process is mostly similar and fluctuating It is due to the synchronous editing between User 1 and User 2 which occurs asynchronously with the server, does not depend on server responses The considered network delay does not affect Typing and Receiving sub-processes

12. Collaborative Scenario – Total Processing Time ( )   increases rapidly at higher delay combined with packet loss It is due to the retransmission of lost packets takes longer time when network has delay The combination of delay and packet loss influence negatively the performance of Google Docs  

13. Linear R egression M odel The increasing of can be modelled using a linear regression for a delay and packet loss : This model has sufficient quality to be used in study purposes  

14. Conclusion The performance of Google Docs was analyzed under the impact of various network parameters in the two scenarios In both scenarios with baseline network delay: Considered packet loss rates do not influence the duration of sub-processes Network delay affect negatively the performance of Google Docs Login and Creating document sub-processes are the most delay-sensitive Considered network delay does not affect the Typing and Receiving sub-processes in collaborative scenario The simultaneous occurrence of delay and packet loss on the network influences significantly on the performance of Google Docs