Quality of Experience in Remote Virtual Desktop Services

Pedro Casas, Michael Seufert,

Sebastian Egger, and Raimund Schatz

FTW Vienna, Austria

Quality of Experience in Remote

Virtual Desktop Services

2 -© FTW 2012

The Cloud paradigm is pushing traditional “local” services to run at

large datacenters in remote locations.

The service quality becomes an important differentiator between ISPs, specially in the case of mobile operators.

Cloud QoE:

- Assess and forecast impacts of migrating services to the cloud

how to maintain good QoE in the Clouds?

- Network dimensioning guidelines for Cloud services (specially at the access network).

The impacts of the Network on Cloud QoE are tied to the

interactivity degree of the application:

Low High Remote Virtual Desktop File Sync and Storage

CRM

Cloud QoE is about interactivity

Tele-presence

The

Project

4 -© FTW 2012

QoE in RVD Study Overview

RVD Traffic Characterization

Results of the Subjective Tests

Conclusions

QoE in RVD Study Overview

RVD Traffic Characterization

Results of the Subjective Tests

Conclusions

6 -© FTW 2012

QoE in Remote Virtual Desktop

The

Approach

8 -© FTW 2012

Testbed Layout using the iLab

Citrix Virtual Remote Desktop System used in the tests

Two identical setups: a laptop and a Citrix Remote Virtual Desktop, provisionned by a

Citrix XenServer (v 6.0.2)

Traffic shaping between the XenServer and the Virtual Desktops

All the traffic packets between the XenServer and the virtual desktops are captured for post-analysis

4 tasks covering the different interaction techniques used in the

context of desktop applications:

- Typing: transcribe a printed text

- Scrolling: document reading with scrolling

- Drag & Drop: drag & drop images to specific screen locations

- Menu browsing: multiples menu-browsing

Gamification approach:

- Implement testing tasks as a game

- Improve participant engagement

- Permits to increase testing duration

- Implementations: Quiz Game, Puzzles, Interactive Menu Browsing

10 -© FTW 2012

Remote Desktop QoE Tests: QoS Levels

Two evaluation phases:

- Lab user tests: QoE for different response times (realized by changing network RTT

– Citrix traffic goes over TCP)

- Mapping response times to QoS requirements (Downlink bandwidth)

Testing conditions:

- 5 QoS levels (RTT = 50, 150, 200, 350, 500 ms)

- Extra QoE level local condition (no RVD)

User experience ratings (by automatic prompting):

- 9-levels ACR scale

- overall experience in this specific task

- perception of network speed

- difficulty in achieving the task

- acceptablility to use the application in the experienced conditions

Participants Demographics

52 participants

Gender: 23 Female / 29 Male

Average age = 32

<30 (28), 30-40 (13), 41+ (11)

Daily Internet usage (hours)

12 -© FTW 2012

QoE in RVD Study Overview

RVD Traffic Characterization

Results of the Subjective Tests

Conclusions

Remote Desktop and Up-link Bandwidth

Uplink traffic throughput consumed by Citrix sessions is negligible

Only user inputs (keyboard, mouse) 100 kbps is enough to transfer

14 -© FTW 2012

Very dependent on the traffic patterns generated by each task

Downlink traffic is determined by the level of interactivity and the screen refreshment rate of the particular task

In the downlink direction, bandwidth has a marked effect on RD QoE

QoE in RVD Study Overview

RVD Traffic Characterization

Results of the Subjective Tests

Conclusions

16 -© FTW 2012

Citrix RVD Response Times

The Citrix Response Time (i.e., time between a user input and the corresponding screen refreshment) is not negligible

Compared to the response times of a local desktop application, Citrix

adds an additional delay of about 150ms, which impacts the QoE of the

RVD QoE vs Network RTT

Depending on their specific characteristics, different tasks have different QoE sensitivity to network impairments.

More interactive and throughput-intensive applications are more sensitive to RTT

RTT should be kept below 150 ms to achieve good QoE and high acceptance

18 -© FTW 2012

RVD QoE vs Down-link Bandwidth

Good QoE can be expected if the downlink bandwidth is high enough to avoid shaping the downlink traffic

A minimum downlink bandwidht of 2 Mbps is necessary to achive good QoE in the evaluated tasks

4 Mbps of downlink bandwidth avoids QoE degradation due to downlink traffic

Impact of RTT on User Behavior and Productivity

Cloud QoE is not only about how smooth the application runs on the client, How difficult it is for a user to interact with a remote system when response times are high

20 -© FTW 2012

QoE in RVD Study Overview

RVD Traffic Characterization

Results of the Subjective Tests

Conclusions

Conclusions of the Study (1/2)

Responsiveness is the key to overall satisfaction with Remote Virtual

Desktop Services (and Cloud QoE in general)

The degree of interactivity of the application is the most

discriminating factor in terms of QoE and influence of the network:

results in RVD show that applications with higher degree of

interactivity are more sensitive to bad network conditions, specifically as regards high network RTT

High RTT values not only translate into bad QoE, but also impact the performance of the end user in terms of the time needed to achieve

specific tasks

Downlink bandwidth should be high enough to avoid downlink traffic shaping so as to avoid QoE degradation

22 -© FTW 2012

Conclusions of the Study (2/2)

Citrix QoE even under optimal network QoS conditions does not reach the QoE of a local desktop, due to an additional response delay of about 150ms introduced by Citrix

Such baseline degradation depends on the specific task (interactivity degree and screen activity), and can be as high as 0.5 MOS for highly interactive graphical tasks (drag & drop of images)

Regarding testing strategies, gamification approaches proved to be very useful to achieve better participation during longer testing