VIMAN lab pushes the virtual network computing envelope
Researchers didn’t have to rub a magic lamp to produce the GENI that is making virtual computer networking dreams come true. GENI is the acronym for Global Environment for Network Innovations. It is a cloud infrastructure sponsored by the National Science Foundation in this country that allows researchers and others to set up virtual labs to share information, resources and data, and to test the possibilities of communication and collaboration on a scale that previously was impossible.
Prasad Calyam, assistant professor in the Department of Computer Science, and the research assistants in his Virtualization, Multimedia and Networking (VIMAN) Lab at the University of Missouri are working at the forefront of new data-intensive application possibilities to test their capabilities and optimize network performance for those accessing and using what essentially is the next iteration of the Internet.
Calyam came to MU from The Ohio State University where he earned his doctorate and also served as research director of the Ohio Supercomputer Center. He and colleagues at MU in collaboration with OSU received NSF funding totaling nearly $2 million to build and closely integrate “Science DMZs” on both campuses, which are 100 gigabit/second secure networks with support for high-volume data movement necessary for scientific research.
Last summer Cisco Systems, Inc. provided Calyam grant funding to develop new technologies to set up dual mode GENI Racks within Science DMZs, one mode for use as a community resource and the other to be used specifically by researchers at MU and OSU.
“It [GENI Rack] is like a server that can be accessed by researchers from other universities, Sripriya Seetharam, a master’s student in Calyam’s lab explained. “GENI serves as an experimental testbed for us to understand network-as-a-service for data-intensive research applications.”
Seetharam and master’s student Ronny Bazan Antequera have been working to set up data-intensive research experiments and classroom lab exercises that utilize GENI Racks and other resources such as data transfer nodes and virtual desktops within the Science DMZs on both campuses.
In the fall, Calyam taught MU’s first cloud computing course utilizing GENI technologies, which covered fundamental high-performance computing concepts that are powering products of companies such as Google, Facebook, Microsoft and Amazon.
The trio published a peer-reviewed paper on planning and outcomes of the course in the NSF-supported GENI Research and Education Experiments Workshop held early this year in NYU-Poly, Brooklyn, NY.
The VIMAN Lab research group is using a couple of different projects to put GENI technologies as well as Science DMZ capabilities and performance to the test. Their goal is to develop and improve resource provisioning and placement algorithms/tools to ultimately optimize user experience when using data-intensive applications. They are utilizing multi-domain monitoring and software-defined networking to do so.
Seetharam said one of the test cases involves an ongoing research project at OSU to improve real-time image processing of neuroblastoma, a malignant cancer of the nervous system that commonly affects infants and children.
“The large-size images generated from a local microscope instrument need to be quickly processed by a remote supercomputer,” Seetharam said of the data-intensive application. “First we must decide what kind of a system and network model we would need, and then we need to automatically provision on-demand resources over the best end-to-end network path available between the two sites over the Internet.”
She said the software-defined networking they are using allows researchers to program and dynamically switch between multiple paths for the best results.
“Medical researchers should be able to click a button to rotate or zoom-in on an image, and it should happen immediately without any perceivable impairments. We are trying to understand how to develop network-as-a-service using such data-intensive applications deeply,” Seetharam added. “This project is a kick start!”
The sheer number of collaborative opportunities at MU that can benefit from Calyam’s networking and cloud computing expertise is one of the things that lured the researcher to relocate to MU from OSU. One of these is a project that he has launched with Professor Marge Skubic from the college’s Department of Electrical and Computer Engineering.
One of Skubic’s aging-in-place, or eldercare, projects involves the use of unobtrusive Kinect sensors developed to manage seniors’ risk of falling. The NSF-funded project on which she and Calyam are collaborating would use GENI-enabled networking and MU’s Science DMZ to connect a physical therapist to clients in their homes to monitor and coach them remotely.
“This project involves a novel interface that supports high-definition video-conferencing between Kansas City partners and physical therapists at MU. It needs a high amount of network bandwidth,” said Manav Singhal, a doctoral student in Calyam’s lab, explaining that the Google Fiber installed in Kansas City will enable the networking research and experimental data collection in the homes of elderly patients.
“We are testing configurations and doing analysis on how the system works to help the therapists and seniors,” he said. “The approaches we are investigating to integrate hybrid cloud technologies to deliver apps can revolutionize computing and broadband.”
Singhal said in the future these applications will be delivered into homes and businesses just as utilities similar to water and electricity are provided.
In addition to the projects he and his team are doing to put GENI technologies and Science DMZs through their paces, Calyam is taking a close look at overall performance. He is one of the world’s leading experts on perfSONAR (PERFormance Service Oriented Network monitoring ARchitecture), a set of infrastructure tools developed through contributions of worldwide teams in order to measure and analyze network performance across multiple domains. He recently chaired an NSF-sponsored workshop on perfSONAR and was also the lead guest editor on related feature topic issues in the Institute of Electrical and Electronics Engineers IEEE Communications magazine, which raised MU’s profile in this area to an international level.
“We don’t yet know where the problems are in network performance,” said Calyam. “The Internet is a network of networks and achieving expected performance is an end-to-end problem.”
His research group is working to develop metrics and adaptive performance sampling techniques to optimize the use of high-speed networks for cloud computing. Quality of service impacts all users, but a problem arises from the fact that commercial companies and network operators do not want to share performance information.
“It is hard to do an analysis without a topology of how the networks are connected, and there are some measurements you just can’t collect and use for analysis and sharing of events such as bottleneck anomalies. It’s hard to build an accurate model with sparse data and use that to alert network operators or users to troubleshoot their poorly performing networks,” Calyam said.
What is needed, he said, is a common policy agreement and innovative technologies to collect, analyze and share diverse measurement resources and data, kind of a Facebook for performance measurements.
The Federal Communications Commission (FCC) also is interested in this work as part of programs such as “Measuring Broadband America” to equip consumers with information about their Internet connection performance and help them make performance fact-based decisions in choosing their Internet service subscriptions.
The Department of Energy has awarded Calyam nearly $500 thousand for basic research to improve perfSONAR. DOE also awarded nearly $1.2 million in Small Business Innovation Research and Small Business Technology Transfer funds to Calyam’s commercialization partner company, Samraksh, to integrate the perfSONAR research outcomes and develop new performance-monitoring software. Calyam has named the software he is developing “Narada Metrics.”
“In Indian mythology, when there is crisis and disagreement between two opponents, Narada [a Hindu sage] collects and shares multi-party information and solves the problem for the greater good. Everyone likes him because he typically arrives when there is little hope to resolve conflict between opponents in a story segment, and his arrival is always a sign of good omen for conflict resolution,” Calyam said.
Calyam said what drives his basic and applied research agenda is that he deeply cares what he does will create new learning opportunities for students and help people.
I want this work to matter to people and to be genuinely useful,” he said.