Ashwin Gumaste’s research while affiliated with Indian Institute of Technology Bombay and other places

Network Function Virtualization (NFV) is a promising technology that can transform how internet service providers deliver their services. However, recent studies have identified several challenges in adopting NFV. Two key challenges are central to the operation and capacity planning of NFV Data Centers (DCs): (i) Service Function Chain (SFC) realization – determining if a new request with a known profile can be accommodated, and (ii) Network Function Virtualization (NFV) traffic bounds estimation – estimating the total traffic that a data center can handle considering all service requests and their performance constraints. To address these challenges, we propose a model that leverages stochastic network calculus to effectively dimension an NFV DC while ensuring delay and availability bounds for all service requests. Our theoretical model provides a mathematical framework to assess the realization of a single SFC request without delving into the specifics of the realization process. We utilize established availability-aware Virtual Network Function (VNF) placement patterns to obtain traffic bounds essential to planning data center capacity. We analyze NFV data center traffic under various scenarios over a Fat-tree DC topology. The results demonstrate that data center capacity is significantly influenced by the VNF placement strategy. Additionally, for data centers hosting latency-sensitive services, Service Level Objective (SLO) constraints on availability and delay are crucial in determining the number of such requests that can be accommodated.

We present Vercel, a network verification and automatic fault rectification tool that is based on a computationally tractable, algorithmically expressive, and mathematically aesthetic domain of linear algebra. Vercel works on abstracting out packet headers into standard basis vectors that are used to create a port-specific forwarding matrix $\mathcal{A}$, representing a set of packet headers/prefixes that a router forwards along a port. By equating this matrix $\mathcal{A}$ and a vector b (that represents the set of all headers under consideration), we are able to apply \textit{least squares} (which produces a column rank agnostic solution) to compute which headers are reachable at the destination. Reachability now simply means evaluating if vector b is in the column space of $\mathcal{A}$, which can efficiently be computed using least squares. Further, the use of vector representation and least squares opens new possibilities for understanding network behavior. For example, we are able to map rules, routing policies, what-if scenarios to the fundamental linear algebraic form, $\mathcal{A}x=b$, as well as determine how to configure forwarding tables appropriately. We show Vercel is faster than the state-of-art such as NetPlumber, Veriflow, APKeep, AP Verifier, when measured over diverse datasets. Vercel is almost as fast as Deltanet, when rules are verified in batches and provides better scalability, expressiveness and memory efficiency. A key highlight of Vercel is that while evaluating for reachability, the tool can incorporate intents, and transform these into auto-configurable table entries, implying a recommendation/correction system.

We present Vercel, a network verification and automatic fault rectification tool that is based on a computationally tractable, algorithmically expressive, and mathematically aesthetic domain of linear algebra. Vercel works on abstracting out packet headers into standard basis vectors that are used to create a port-specific forwarding matrix $\mathcal{A}$ , representing a set of packet headers/prefixes that a router forwards along a port. By equating this matrix $\mathcal{A}$ and a vector b (that represents the set of all headers under consideration), we are able to apply least squares (which produces a column rank agnostic solution) to compute which headers are reachable at the destination. Reachability now simply means evaluating if vector b is in the column space of $\mathcal{A}$ , which can efficiently be computed using least squares. Further, the use of vector representation and least squares opens new possibilities for understanding network behavior. For example, we are able to map rules, routing policies, what-if scenarios to the fundamental linear algebraic form, $\mathcal{A}x=b$ , as well as determine how to configure forwarding tables appropriately. We show Vercel is faster than the state-of-art such as NetPlumber, Veriflow, APKeep, AP Verifier, when measured over diverse datasets. Vercel is almost as fast as Deltanet, when rules are verified in batches and provides better scalability, expressiveness and memory efficiency. A key highlight of Vercel is that while evaluating for reachability, the tool can incorporate intents, and transform these into auto-configurable table entries, implying a recommendation/correction system.

We examine how to provide applications with dedicated bandwidth and guaranteed latency in a programmable mission-critical network. Unlike other SDN approaches such as B4 or SWAN, our system Tuneman optimizes both routes and packet schedules at each node to provide flows with sub-second bandwidth changes. Tuneman uses node-level optimization to compute node schedules in a slotted switch, and does dynamic routing using a search procedure with QoS-based weights. This allows Tuneman to provide an efficient solution for mission-critical networks that have stringent QoS requirements. We evaluate Tuneman on a telesurgery network using a switch prototype built using FPGAs, and also via simulations on India’s Tata Network. For mission-critical networks with multiple QoS levels, Tuneman has comparable or better utilization than SWAN while providing delay bounds guarantees.

5G promises to be revolutionary from the perspective of offering unprecedented bandwidth to the end user. In this pursuit, the limited spectrum of the wireless overlay, further constricted by distance, requires strong support from a backhaul network. Such a strong backhaul network should provide large bandwidth pipes with near proximity to the end user at low costs and using pragmatic technology. Naturally, an optical solution is best suited to meet the voluminous bandwidth requirements of the backhaul. This optical solution has to cater to dynamic bandwidth needs among base stations and provide rapid and smart provisioning to meet the requirement of new features such as Coordinated Multipoint (CoMP) among base stations, and yet be low cost. To this end, we propose the use of light-trails (LTs), which are a generalization of a lightpath that provides for dynamic bandwidth communication, sub-wavelength optical grooming and optical layer multicasting using contemporary optics. We observe that each feature of a LT, i.e., dynamic provisioning, optical multicast, and sub-wavelength grooming is strongly desired by the 5G backhaul, implying a clear mapping between them. We investigate the usage of light-trails as a 5G backhaul solution. We discuss the proposed architecture of light-trails for micro cells, pico cells, and femto cells. We present the LT design problem for 5G backhaul as a constrained optimization problem. We also present a mechanism to achieve coordinated multipoint using light-trails. A dynamic virtual topology growth algorithm for the backhaul is also presented and extensively evaluated. The light-trails approach is compared to other approaches, the use of an SDN controller is discussed and significant performance benefits are observed.

Background: . In this work, we explore the possibility of decoding Imagined Speech (IS) brain waves using machine learning techniques. Approach: . We design two finite state machines to create an interface for controlling a computer system using an IS-based brain-computer interface. To decode IS signals, we propose a covariance matrix of Electroencephalogram channels as input features, covariance matrices projection to tangent space for obtaining vectors from matrices, principal component analysis for dimension reduction of vectors, an artificial neural network (ANN) as a classification model, and bootstrap aggregation for creating an ensemble of ANN models. Result: . Based on these findings, we are first to use an IS-based system to operate a computer and obtain an information transfer rate of 21-bits-per-minute. The proposed approach can decode the IS signal with a mean classification accuracy of 85% on classifying one long vs. short word. Our proposed approach can also differentiate between IS and rest state brain signals with a mean classification accuracy of 94%. COMPARISON: . After comparison, we show that our approach performs equivalent to the state-of-the-art approach (SOTA) on decoding long vs. short word classification task. We also show that the proposed method outperforms SOTA significantly on decoding three short words and vowels with an average margin of 11% and 9%, respectively. Conclusion: . These results show that the proposed approach can decode a wide variety of IS signals and is practically applicable in a real-time environment.

Grafnet, a Graph Neural-Network (GNN)-based scheme learns IP-address-to-port mapping, leading to forwarding table-less routers. GNNs allow mapping network-wide features like adjacencies and addresses to generate new representations. Grafnet converts network-wide IP addresses to a feature space using GNNs. GNNs extrapolate node adjacencies onto a feature matrix, whose output tells which address/subnet is connected to a node and port. To do so, we use a GNN in conjunction with an Artificial Neural Network (ANN), whose output transforms graph adjacencies to address-based adjacencies. We exploit the fact that IP addresses are present in contiguous groups (subnets) or ‘ranges’. Large range sizes imply a better likelihood of Grafnets’ approximation, though with enough learning Grafnet learns just about all network-wide IP addresses, irrespective of range sizes. Grafnet is evaluated as an SDN scheme on (1) 75-node US-core network and (2) 2000-node, 5 million IP address-based random WAN topology. Analytically, we show equivalence between Grafnet and a Feed-forward neural network implying exhaustiveness and correctness. The proposed Grafnet model is able to work as a direct address translator without the need for tables in the forwarding plane of a router. Engineering considerations to implement Grafnet are also discussed.