Title: The digital wave in communication: re-imagining the future
Speaker: K. Ananth Krishnan, CTO, TCS
Scribe: Partha Sarathi Paul
The change in consumer behavior in this digital era led to steep decline in the telecommunication company (telco) revenues. The speaker highlighted the need to acknowledge the paradigm shift and necessity to change the business direction to keep up with the consumer demands. The application of Artificial Intelligence (AI) techniques in the field of telecommunications is one of the directions to go.
Telecommunications; media and entertainment
The availability of massive computing power and increased interest in multi-disciplinary collaboration blur the traditional boundaries that existed between different domains. For example, the development of autonomous vehicles involves collaboration of trained experts in disciplines such as electrical and electronics, AI, computer vision, etc. In the case of media and entertainment, consumers are increasingly embracing cutting-edge gadgets to consume contents. For instance, there are millions of viewers all over the world who followed almost all of the 2012 London Olympic Events via iPad. This is in contrast to viewership pattern of 2008 Beijing Olympics who stuck to traditional transmission medium such as television.
Some possible directions that can be leveraged by the telecommunication industry to reinvent itself:
Try to deliver seamless and enriched customer experience through comprehensive range of digital touch points
Empower digital enterprises. Act as an enabler of industries such as banking, insurance, etc.
Act as a value integrator in digital ecosystem. Play the role of intelligent hub between a pool of otherwise uncorrelated industries and their customer and generate revenue
Various fields from academia now collaborate with various other fields of science resulting in the birth of new interdisciplinary areas. In this light, AI is gaining new importance with applications in almost all traditional and emerging interdisciplinary areas of interest. The speaker cited an experiment conducted at University of Maryland, USA where a robot is trained to cook food by training the robot using tens of cooking-related videos from YouTube. AI also acts an enabler in various areas such as:
Parallel processing: faster processing of GPU chips inspired by information processing in the brain
Analytics: big data analytics is of great help for brain-like processing
The telecommunication industry can leverage the massive volume of customer call list to generate some useful information. AI techniques can be used to analyze properties of the networks by decoupling information from the source. AI also help in the field of cognitive IoT where a system can understand its goal, state, and the context. AI can be applied for fraud detection by analyzing user behavior data from information sources such as event data record. AI research can also benefit the BPO industry where techniques can be used for voice analysis, non-linguistic speech processing, automatic accent analysis, and emotion detection in real-time.
Q: Will robotic cook ever fully replace human cook in Indian scenario?
A: Possibly not. But there are other areas such as content generation where AI could lead to negligible need for human intervention. Example: a newspaper edition fully generated by AI agent(s).
Q: What is the most critical issue in the field of IoT?
A: Identification of business model that generate revenues besides classic technological challenges such as addressing scalability and security issues and so on.
Session T1-3: Cognitive Radio Networks
Scribe: Akrati Saxena
Title: k-Channel Connected Topology Control Algorithm for Cognitive Radio Networks
As we know there is a large under-utilized portion of the licensed spectrum in spatio-temporal domains. The current paper aims to solve the connectivity problem in Cognitive Radio Networks (CRN) considering the activities of the primary users (PUs).
Cognitive Radio Networks
Authors aim to control the topology through channel assignment to achieve conflict free k-channel connected topology in CRN.
A centralized and power efficient algorithm to achieve k-channel connectivity with conflict free property has been proposed
An upper bound on the number of channels required for the same is specified.
The proposed approach needs a large amount of dynamic data that is not easy to collect.
As authors also explained, a distributed algorithm can be designed to achieve k-channel connected and conflict free topology.
Title: Spectrum Sharing: How Much to Give
The aim of this work is to improve the efficiency of spectrum allocation. For example in India spectrum for cellular technology is allocated in 6 hands and there are several service providers. To address this problem, this present work advocates spectrum sharing.
Spectrum allocation and efficient spectrum usage
The authors introduce partial spectrum sharing approach using which two or more service providers share spectrum resources to improve resource usage while reducing congestion. The authors also propose two different models, namely: 1. Probabilistic Model and 2. Deterministic Model for partial spectrum sharing.
It is a novel technique and can be implemented in India to maximize resource usage
Blocking probability has also been considered
Using the proposed method, price of a free channel can be dynamically decided
Implementation time in developing countries such as India could be longer
Title: Characterization of the Missed Spectrum-Access Opportunities under Dynamic Spectrum Sharing
This paper focuses on quantifying spectrum access opportunities under dynamic spectrum sharing. It also aims to characterize spectrum usage.
Spectrum sharing in CRN
The authors study the effect of lack of knowledge of the propagation environment and the PU receiver positions.
The work emphasizes the case of dynamic spectrum sharing
Authors use MUSE methodology to characterize and quantify the missed spectrum access opportunities
They have shown the experimental results for the proposed idea.
Signal data collected in real-time has been used to understand dynamic spectrum sharing.
Static and exclusive spectrum allocation results in under-utilization of spectrum at most of the locations and times. So the proposed method will also help understand the utilization aspects of spectrum sharing.
It focuses only on dense spectrum sharing which requires complete knowledge of the network and its dynamics
Proposed method uses many dynamic parameters like information of transmitter and receiver and so on.
Title: Cognitive Algorithms for LTE Cloud-RAN
Paper aims to do the joint scheduling in real time
Joint Scheduling in real time
Utility of Joint Scheduling: using fluid queues to establish the benefits of joint scheduling.
Real time Joint Scheduling Algorithm: heuristic real time joint resource allocation algorithm.
Measurement Optimization: link clustering technique to reduce uplink overhead
Simulation results on real-time joint scheduling has been presented
Scheduling algorithm is 45 % energy efficient with respect to baseline
There is a scope to improve the efficiency of the proposed approach
More real-time parameters can be considered
Title: Energy efficient rate coverage with base station switching and load sharing in cellular networks
Optimize minimum possible energy consumption in the case of independent BSS
Cooperation between multiple operators for BSS
The authors derive the rate coverage expression for a scenario where multiple operators are cooperating. They also propose practical SLA that ensures fairness among the cooperating operators and reduce the problem to real-time.
This method can be used to efficiently utilize the spectrum.
Session T2-3: Cloud Computing and Applications
Scribe: Anand Madhavrao Baswade
Title: Per-VM Page Cache Partitioning for Cloud Computing Platforms
To design and implement a scalable general-purpose high-performance page cache solution for the cloud
Linux cache partitioning, Linux-based virtualization
The work advocates allocation of physical resources such that the hypervisor ensures that the VMs do not interfere with each other and can run in isolation, as if each VM has a dedicated hardware underneath.
The authors propose a deterministic page cache on a per-VM basis
The proposed approach would allow us to consolidate more VMs per physical server.
Improved application performance and reduce system load.
A novel hit-rate based cache partitioning based on hit rate curve.
High search cost in the event of cache miss
Title: Implementation of Virtual Sensors for Building a Sensor-Cloud Environment
Influence the use of small sensors
Integrate them with cloud and IoT
Virtual sensors to support e-health and environment monitoring.
The authors propose to create and include virtual sensors, i.e. abstractions of physical sensors, in cloud that can enable on demand, pervasive, shared access to sensors.
The paper presents implementation of virtual sensors at IaaS level.
This paper provides a virtual sensor system architecture for both health and environment application domains and address the shortcomings of existing physical sensors
It provides description of APIs for sensing, processing, communication and storage at IaaS end.
Title: Benchmarks to compare cloud service providers for seamless customer migration
To come up with a mechanism to compare different cloud service providers using different service quality attributes. It would help customers choose a service provider that better suit their requirements when they want to switch from one service provider to another.
Quality of service and SLA assessment in cloud environments
The authors proposed a novel measurement approach wherein the easy-to-measure SaaS layer attributes is used to infer hard-to-measure infrastructure parameters. The measurement approach is non-intrusive, stealthy, and on-the-fly.
The measurement approach sheds some insight into the business model of the infrastructure service provider
Helps identify whether the service provider complies with the guarantees specified in the SLA
Cross-layer mapping for measuring service quality
Identifying non-compliance is much more complex than what is being considered in the work. There could be several set of parameters contribute to service quality degradation rather than simple reduction in VM cycles and bandwidth.
Assumptions such as 100% trustworthy SaaS provider and well-behaved client traffic may not hold in a real-world scenario.
Title: cStream: Cloud based high performance video delivery network
Live streaming and video-on demand are increasing at a rapid pace.
Global Over-the-top (OTT) video market is estimated to grow to $37.2 billion by 2017.
Cloud-based video streaming
cStream leverages the Cloud provider’s global footprint, Cloud provider’s high performance backbone network between different data centres, social media analytics and a UDP based fast data transfer protocol to optimize the quality of experience for end users, and the total cost incurred by the cloud provider in terms of network bandwidth and compute resources
The paper presented the design and implementation of cStream - a high performance cloud based live video delivery network that leverages the Cloud provider’s global footprint, Cloud provider’s high performance backbone network between different data centres, social media analytics and a UDP based fast data transfer protocol.
cStream can improve throughput and transfer times by up to 14 times for a transmission across the globe from San Jose to Hong Kong
Title: Cost-aware Capacity Provisioning for Fault-tolerant Geo-distributed Data Centres
Need for cost-aware capacity provisioning for geo-distributed data centres that can tolerate failure of a single data centre.
Data centre fault tolerance
The paper proposed a Mixed Integer Linear Programming (MILP) optimization model that reduces the total cost of ownership (TCO) (includes capital and operating cost factors) while provisioning the servers across different locations with varying running cost factors.
The paper propose an optimization framework to minimize the Total Cost of Ownership (TCO) of the cloud provider while designing fault-tolerant geo-distributed data centres.
The paper considers demand, delay, compute capacity, power availability locations in optimization problem and showed results for power usage effectiveness (PUE), over-provisioning for fault tolerance, and choice of data centre locations, and latency requirements on the TCO.
Title: Leveraging the Trade-off Between Spatial Reuse and Channel Contention in Wireless Mesh Networks
Speaker: Subhendru Chattopadhyay (IIT, Guwahati,India)
Scribe: Alok Ranjan
Motivation: Wireless communication technology is witnessing rapid changes in the way the network deployment is increasing and being used by the people. The main goal of this work is to have better throughput without depending on the transmit power, data rate and scheduling.
Application Areas: Telecommunication Industries, Cellular networks, Wireless Communication
Description: Current state of arts in the wireless communication area is changing rapidly from spectrum sensing, IoT, 5G, cognitive radio and etc. However, there are few concerns i.e. the end to end delay and the spectrum reuse. These needed to be considered for the real time deployment. Therefore, a trade-off is needed between the spatial reuse and channel contention. The present work proposes a solution based on the simulation analysis considering different parameters. The proposed solution is based on the distributed heuristic approach.
Proposed system model considerations:
Wireless mesh networks, IEEE 802.11 b/g/n, physical Layer Support, SDR supported, single interference and single channel.
Formalization of problem (solution approach):
Connectivity matrix, antena and channel gain.
Constraints to the proposed solution which have been considered for distributed system:
Hidden node, SNR.
The proposed model is based on three key parameters: Jain's fairness index, end to end throughput
and end to end delay of the networks.
The system provides improved Fair JPRS. The required average power level and throughput remains almost similar.
Pros of the talk:
New approach to solve the end to end throughput for the wireless mesh networks.
Improved performance compared to the existing one.
Cons of the talk:
Work is based on the simulation, no experimental data
Title: An Improved Multiple Feedback Successive Interference Cancellation Algorithm for MIMO Detection
Scribe: Prateek Khandelwal
The paper addresses the symbol vector detection problem in MIMO systems. It also increases the channel capacity and detects of multiple data streams, reduces the computational effort and increases the efficiency
Physical layer communications and large MIMO systems.
The paper controls the error propagation in successive interference cancellation. This is achieved using recursively checking for unreliability condition (IMF-SIC) and multibranch technique with different orderings (MB-IMF-SIC).
The paper reduces error rate and it is closer to ML’s performance. Multibranch approach leads to near optimal performance.
IMF-SIC’s tree for larger number of antenna leads to widening of the tree. Hence it causes non optimal solutions and increased complexity.
Title: Cumulant Based Automatic Modulation Classification of QPSK, OQPSK, 8-PSK and 16-PSK.
Scribe: Prateek Khandelwal
Cumulant based technique are resistance to noise.
Military applications and cognitive networks.
Classification between QPSK, OQPSK, 8-PSK and 16-PSK. Discrimination between 8PSK and 16PSK using squared signal, and between OQPSK 8-PSK
Low SNR ratio can be achieved, along with perfect classification. This is also robust against constellation rotation.
Title: Design of Compact Bent Dipole Antenna and its Array with High Gain performance for GPS Application
Scribe: Prateek Khandelwal
The goal of this paper is increasing the gain for GPS applications. The GPS fails to recieve signals in some environments. Also, the printed antennas are light weight, compact and easily integratable.
Physical Layer Communication, GPS devices/modules.
Coming up with better dimensions for bent dipole antenna. Simulation based evaluation of different configuration of antenna.
The proposed approach optimizes on the metric of return loss and high gains.
Scribe for Panel Discussion
Net neutrality: Is it good, bad, or confusing for the telecom industry?
Scribe: Anamika Chhabra , Karan Grover
Trends that have recently fired up the debate of net neutrality are the zero rating schemes introduced by Airtel, reliance and other telecom operators. Also the free basics scheme introduced by Facebook’s internet.org
General public welfare, Upliftment of the society, More literacy and awareness
Main discussion points:
The topic of net neutrality is being discussed a lot lately on the news and social media. Net Neutrality means an Internet that enables and protects free speech. It means that Internet service providers should provide us with open network and should not block or discriminate against any applications or content that ride over those networks. Just as your phone company shouldn't decide who you can call and what you say on that call, your ISP shouldn't be concerned with the content you view or post online. Without Net Neutrality, cable and phone companies could carve the Internet into fast and slow lanes. An ISP could slow down its competitors' content or block political opinions it disagreed with. ISPs could charge extra fees to the few content companies that could afford to pay for preferential treatment thus relegating everyone else to a slower tier of service. This would destroy the open Internet. There are many aspects to the “net neutrality” that should be taken into consideration if correct policies have to be made around it. These are civil society perspective, economic and investment in content.
Most discussion sas we are seeing today surround the civil society perspective. The fact that is most stressed upon is that every individual should have equal access to the internet, especially the poor. The economic perspective and the investment perspective are the least stressed upon and taken into account on the effects of competition in the market etc. Some have also argued that the champions of net neutrality are fighting for something which never was. The Internet in its whole, seems elegant from the above but is actually very complex from below. Many economic entities play a role in the transfer of a single packet and it is quite obvious that the prioritization based on the content almost always happens, even if in the most subtle way.
The panel discussed four types of solutions that can be implemented. Those are different permutations and combinations of whether there should be differential pricing and/or differential priority. Many people seemed to agree with either one of the two being implemented. One Major suggestion is to do more innovation on business models. Because the zero rating model is not bringing money to telecoms, it is taking money away from them. Hence that business model does not seem to be sustainable.
Coming back to the civil society perspective, it was argued that India has about 20% Internet presentation. 80% of the country does not have Internet access and a major portion of that number consists of poor people who cannot afford it. It therefore does not make sense to keep them away from the entire summation of human knowledge which can change their lives for the better because of a theoretical concept like net neutrality. But it should not be forgotten that mobile connectivity is still in its infancy in India, despite its wide reach. Internet connectivity is very affordable in India compared to other countries, but still a lot has to be done.
Question and answers:
Question: Instead of looking for solutions in policies and regulations, is it possible that solution lies in the marketplace? May be what we need is a solution that compels people to get online?
Answer: Some kinds of policies are necessary to be put in place. However, there are actually two dimensions, which can be looked at: 1.) differential pricing and 2.) differential priority. In order to provide a solution, we cannot have both. We want some flexibility, we can either have differential pricing with no priority or we can have differential priority with no pricing.
Question: What is the importance of Net neutrality in the context of Indian environment?
Answer: In India, it is perceived as: If you are not neutral, then you are not poor! Here we have a big population who does not even have a smart phone, so before looking into net neutrality aspect, it is important that people should be exposed to proper infrastructure for actually availing the benefits of net neutrality.
The research direction is essentially in two directions: business and technology.
One has to work on different business models that are both innovative, sustainable and achievable with the objective of getting as many people online as possible.
The other is to get over the technical challenges of municipal WiFi and other challenges.