VDPilot: Feasibility Study of Hosting Virtual Desktops for Classroom Labs within a Federated University System Presenter: Prof. Jay Ramnathan Authors: Prasad Calyam, Alex Berryman, David Welling, Saravanan Mohan, Rajiv Ramnath**, Jay Ramnathan** Ohio Supercomputer Center/OARnet, **Dept. of Computer Science and Engineering, The Ohio State University Point-of-Contact: pcalyam@oar.net IBM Cloud Academy Conference, April 2012 Topics of Discussion • • • • • • • Motivations for Virtual Classroom Labs Project Overview VDPilot Testbed Registered Users and Workflow Subjective Testing Methodology and Results Lessons Learned; “User Quotes” Next Steps 2 Motivations for Virtual Classroom Labs • Access to expensive, computational software such as SPSS and Matlab is a logistical and licensing challenge – Students: need pervasive access to the software, and have trouble obtaining a license and installing the software correctly on their home computers – Professors: want to manage lab exercises, assignments and exams, and are led to using e-mail to send and receive large files, and are limited in their ability to access and assist in the work-in-progress of the students – CIOs: want to use federated shared infrastructure resources that would simplify classroom lab computing for faculty and students, and reduce costs for universities 3 VDPilot Project Overview • VDPilot project initiated by BOR-CIO Advisory Council – Extends our ongoing VMLab efforts – http://vmlab.oar.net • Built through collaborations with IBM, VMware, Dell, NSF • Feasibility study of a “virtual desktop cloud” for classroom labs – Leverages the universities’ access to high-speed OARnet network and infrastructure – Expected outcomes for universities: (i) assessment of user experience, and (ii) analyze cost savings, challenges - due to shared resources • Testing both open-source VCL and commercial VDI solutions – ~50 users supported in Phase-1; each user given access to a VCL and VDI virtual desktop to compare “going” to a lab versus remote access • Users were Faculty and Students within OARnet members, IT Admins – Exemplar applications: MS Office, Matlab, SPSS, Windows Media Player, Internet Explorer – all on Windows 7 platform • VDPilot Website – http://vdpilot.oar.net – User testing for Phase-1 - Sept. 22nd till December 8th 2011 4 VDPilot Testbed 5 Registered Users • Total Users Registered: 36 • Participating Institutions: 11 – OSU, U Dayton, U Akron, Ohio U, Walsh, Denison, BW, Sinclair, Ashland, OARnet, OSC • Subjective Testing Status: (September 22nd 2011 – December 8th 2011) – 38% have completed testing – 34% have partially completed testing – 28% are yet to start testing 6 User Workflow • User registers at VDPilot website and requests VCL/VDI virtual desktops • User receives an email from OARnet with virtual desktop login information, testing instructions and link to a survey • Testing instructions guide user through subjective and objective testing of virtual desktop user experience – Subjective testing with application examples • After testing ends, user fills out a “Survey” on VDPilot website – Connection details (wired/wireless, client OS), Subjective Opinion Scores, Other feedback, … 7 Illustrative Screenshots on Testing Project Testing Introduction 8 Illustrative Screenshots on Testing (2) Subjective Testing Manuals Testing Survey Form 9 Subjective Testing Methodology • Participants evaluate their user experience for test applications: – Excel, SPSS, Matlab, Windows Media Player, and Internet Explorer • Application-specific instructions: – Excel – “Please create sample data and then create a chart for the data. While using Excel, pay close attention to the response time of your actions.” – Windows Media Player – “Please play a video from the virtual desktop's hard drive and observe the audio/video quality of the video.” – Internet Explorer – “Please go to http://www.cnn.com and observe how well the web page performs.” – Matlab – “Please load a pre-created script which creates a horse model and then manipulate the model in the graph viewer. You should focus on the time it takes to create the graph, how responsive the graph tools are and overall response times while using the program.” – SPSS – “Please load pre-created data, add an entry, and run a pre-created script to analyze the data; focus on how well SPSS runs, and how quickly issued commands were processed.” 10 Subjective Testing Results • User environment (OS, Network connection) Data – Most used Windows 7, one used Windows XP – Most used wired, couple of wireless last-mile connections • VDI/VCL Preference Results – Both VDI/VCL evenly liked, some say they will use either 11 Subjective Testing Results (2) • Preference comparison of “virtual” with “physical at home” – Most liked using virtual – Some preferred home better, some even preferred virtual better!, some could not tell the difference 12 Preference of VDI/VCL for Apps Users preferring VDI • Users preferring VDI, scored all app user experience at high levels Users preferring VCL • Users preferring VCL, scored app user experience over a wider range 13 Lessons Learned • Well-provisioned networks are key for usability – OARnet regional-network advantage for universities • “Open” versus “Proprietary” solution trade-offs – Setup, maintenance, cost • Licensing challenges – Vendor support models in virtual environments is emerging • Economic viability – Smaller organizations need is greater • Dire need exists in Education community – E.g., Two Professors inquired whether their students could use the VDPilot virtual desktops in their course immediately! 14 “User Quotes” • Went in and completed it this morning and finished up the subjective survey. 2 very nice products. • Word works better than I expected; I think it (VCL) is a better fit for us at Denison than VMware View. I believe it would allow for the use of physical systems in labs that are locked after hours . Thanks for setting this up. • It (virtual) was a bit slower (compare to physical at home), but it would be great for students to have access to programs like SPSS from houses and dorm rooms, since downloading it to laptops proves to be a challenge every semester. • Very similiar (virtual compared to physical desktop at home); SPSS runs slow on the university computers too, and this virtual SPSS performed as well as the computers in the lab. 15 “User Quotes” (2) • I would prefer VDI if the desktop was persistent so I could continue working on the same desktop from anywhere. • Both worked equally well on my PC. • I liked the fact that VCL is open source; the VCL setup seems like it could be a decent substitute for my home machine. • The testing process went well, documentation was easy to follow, and only took me twenty minutes. 16 Next Steps • Discuss with application vendors regarding licensing – Realize “Pay-as-you-go” licensing in virtual environments • Increase the number of participant institutions – Involve wider demographic and underserved institutions • Add federated authentication for virtual desktop access – Investigate Shibboleth-like frameworks • Re-run pilot with a distributed data center architecture – Serve a variety of institutions across the state of Ohio – Investigate power-savings, migration of virtual desktops, … 17 Thank you for your attention! 18