Bookmark and Share


Networking Systems Department

Computing Technologies Department


 Networking Systems Department: 


1) VRAN – Enabling a Flexible and Programmable Cellular Infrastructure

Traffic predictions show that the user demand will continue to increase exponentially in the coming years. Wireless data demand will grow fastest, projected to be a factor of 18 over the next 5 years. In order to address these demands, Wireless networks need to become more flexible and programmable. Currently, operators are required to make significant investments in new infrastructure and infrastructure upgrades to bring new innovative solutions and standard releases to the market. Future network infrastructure needs to be able to quickly offer new types of services in a dynamic way, such as M2M communication and the Kindle model. Furthermore, network providers are interested in being able to share infrastructure as this enables new business models in which infrastructure providers are separated from service providers. This is a model that is already emerging today, for example, lightSquared had offered a wholesale LTE network to MVNOs and operators.

For cellular networks, this can be achieved by virtualizing the radio access network such that all radio processing can be performed in software. Building a virtualized radio access network (VRAN) requires significant new research. A first challenge is the ability to virtualize and dynamically instantiate and run multiple base stations (possibly of different standards) on a common (possibly heterogeneous) programmable platform. This is especially challenging since wireless processing is extremely resource intensive and has stringent real time requirements. The problems posed here are significantly different from traditional real time systems and requires the development of new resource management algorithms. The second challenge is to scale processing to data center dimensions where the compute resources are comprised of pools of possibly heterogeneous resources. This requires developing an “operating system” that abstracts these heterogeneous resource pools as a common shared resource and instantiates and manages them dynamically as the processing load varies.


2) Moving enterprise networks and applications to cloud

This project involves creating systems and network support for moving various enterprise applications and network appliances to the cloud. The research problems consist of how to create right system abstractions for running applications in the cloud, how to virtualize various network elements like routers and middle boxes and the network as a whole, and optimization algorithms for resource placement in these new settings. For moving legacy applications to the cloud and scaling up and down as needed, we create a unified view of resources that combine resources from multiple virtual machines (VMs) and expose it to the application as a single resource. We also create a unified network view for these VMs using network virtualization. For providing enterprise network services in the cloud, we create instances of virtual network locations in the cloud that connects to enterprise using simple customer premise equipments (CPE). We also create very scalable, reliable instances of different network appliances (middleboxes) by re-architecting and separating compute, lookup and storage elements in these appliances. We also work on optimization algorithms for efficient placement of resources in cloud settings like for both Map/Reduce and traditional jobs.


 Computing Technologies Department:


1) Algorithms and Machine Learning

a) Social Network Analysis: Application of machine learning algorithms to social network analysis in order to explain social phenomena like influence propagation, community formation, social collaboration etc. The scope of this research direction is broad - at one end, we are keen on analyzing social network data from online social networks (OSNs e.g. Twitter), or collaboration networks (DBLP, GitHub etc.) to develop probabilistic models to learn and characterize these social phenomena. On the other hand, we are interested in collating and mining data from disparate sources (network data from routers, base-stations or wifi access points and social content from OSNs) to build joint models for predicting social network activities and network traffic.
Finally, these learning algorithms need to be distributed to make inference scalable on large data sets.

b) Privacy: Intelligent systems that learn user privacy preferences (by analyzing web browsing history and social communication logs), and detect, notify and sanitize privacy breaches in real-time.

c) Smart Mobile Systems: We are looking to build intelligent and interactive application platforms on smart-phones. Depending on the goals, this requires the development of algorithms that can learn & process multi-modal inputs (voice, video via a camera, location, text etc.) and through interactions with the end-user, offer fast and accurate decisions within the resource constraints of a mobile device and the wireless network.

Researchers: Anirban Majumder, Nisheeth Shrivastava, Samik Datta, Sharad Jaiswal, Sreedal Menon 




2) Privacy preserving personalization

The P3 project at Bell Labs India (in collaboration with Bell Labs France) is focused on addressing a very pressing problem afflicting Web users of today - i.e. how to safeguard their privacy from the ‘Big Brothers’. End-users of a broad class of applications (e.g. Google services, Foursquare, Facebook) suffer from the dilemma of having the end-user disclose sensitive profile information (browsing history, location tracks etc) in order to benefit from personalized content/services.

P3 is a distributed cloud-based middleware (similar to a distributed set of proxies) that interfaces between end-users and content providers, and enables end-users to avail of recommendation services without having to give up their sensitive profile to these content providers, thereby safeguarding the privacy of end-users. P3 is designed to support several classes of applications – Location Based Services (LBS), Recommender systems, and Online Social Networks (OSNs). Additionally, inspired by the seamless interfacing approach of CDN architectures like Akamai, P3’s design goal aims for seamless integration with existing content provider frameworks (e.g. Google), thereby resulting in large-scale adoption without disrupting the existing application ecosystem.

Researchers: Animesh Nandi, Sharad Jaiswal


3) Technologies for Emerging Markets

a) CommunityWeb: While content-creation tools like blogs and YouTube have resulted in explosion of user-generated content on the web, their adoption has been limited at the grassroots in emerging markets. We are working with the Community Radio ecosystem in India to develop a multimodal web based content creation and dissemination platform for the grassroots.

b) Security protocols for branchless banking: Branchless banking systems offer the possibility of access to banking services to half the world’s population without a bank account. These systems typically involve a human agents (usually, local community members) to mediate transactions, and mobile technologies for remote communications. In our work, we are exploring novel mechanisms for user authentication and transaction receipt confirmation, which are secure, yet easy to use. These mechanisms are necessitated by constraints such as lack of programmable phones, new threats posed by human intermediaries, and cost and usability concerns.

Researchers: Akhil Mathur, Saurabh Panjwani, Sharad Jaiswal