Mohit Tawarmalani

hear(pronounce)

Refereed Conference Proceedings

  1. Nagoja, A., Tawarmalani, M., & Agrawal, R. (2024). An MINLP formulation for global optimization of heat integration-heat pump assisted distillations. Proceedings of the 10th International Conference on Foundations of Computer-Aided Process Design.

    AbstractThermal separation processes, such as distillation, play a pivotal role in the chemical and petrochemical sectors, constituting a substantial portion of the industrial energy consumption. Consequently, owing to their huge application scales, these processes contribute significantly to greenhouse gas (GHG) emissions. Decarbonizing distillation units could mitigate carbon emissions substantially. Heat Pumps (HP), that recycle lower quality heat from the condenser to the reboiler by electric work present a unique opportunity to electrify distillation systems. In this research we try to answer the following question in the context of multi-component distillation – Do HPs actually reduce the effective fuel consumption or just merely shift the fuel demand from chemical industry to the power plant? If they do, what strategies consume minimum energy? To address these inquiries, we construct various simplified surrogate and shortcut models designed to effectively encapsulate the fundamental physics of the system. These models are integrated into a superstructure-based Mixed-Integer Nonlinear Programming (MINLP) formulation, which is amenable to global optimization algorithms aimed at minimizing the effective fuel consumption of the system. Moreover, through the examination of a toy 4-component alcohol separation example, we demonstrate how HPs can notably reduce carbon emissions, even when the consumed electricity is generated by burning fossil fuels.

  2. AlQiam, A. A., Yao, Y., Wang, Z., Ahuja, S. S., Zhang, Y., Rao, S. G., Ribeiro, B., & Tawarmalani, M. (2024). Transferable neural WAN TE for changing topologies. Proceedings of the ACM SIGCOMM 2024 Conference, 86–102.

    AbstractRecently, researchers have proposed ML-driven traffic engineering (TE) schemes where a neural network model is used to produce TE decisions in lieu of conventional optimization solvers. Unfortunately existing ML-based TE schemes are not explicitly designed to be robust to topology changes that may occur due to WAN evolution, failures or planned maintenance. In this paper, we present HARP, a neural model for TE explicitly capable of handling variations in topology including those not observed in training. HARP is designed with two principles in mind: (i) ensure invariances to natural input transformations (e.g., permutations of node ids, tunnel reordering); and (ii) align neural architecture to the optimization model. Evaluations on a multi-week dataset of a large private WAN show HARP achieves an MLU at most 11% higher than optimal over 98% of the time despite encountering significantly different topologies in testing relative to training data. Further, comparisons with state-of-the-art ML-based TE schemes indicate the importance of the mechanisms introduced by HARP to handle topology variability. Finally, when predicted traffic matrices are provided, HARP outperforms classic optimization solvers achieving a median reduction in MLU of 5 to 10% on the true traffic matrix.

  3. Jafri, S. U., Rao, S., Shrivastav, V., & Tawarmalani, M. (2024). Leo: Online ML-based traffic classification at Multi-Terabit line rate. 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI 24), 1573–1591. URL

    AbstractOnline traffic classification enables critical applications such as network intrusion detection and prevention, providing Quality-of-Service, and real-time IoT analytics. However, with increasing network speeds, it has become extremely challenging to analyze and classify traffic online. In this paper, we present Leo, a system for online traffic classification at multi-terabit line rates. At its core, Leo implements an online machine learning (ML) model for traffic classification, namely the decision tree, in the network switch’s data plane. Leo’s design is fast (can classify packets at switch’s line rate), scalable (can automatically select a resource-efficient design for the class of decision tree models a user wants to support), and runtime programmable (the model can be updated on-the-fly without switch downtime), while achieving high model accuracy. We implement Leo on top of Intel Tofino switches. Our evaluations show that Leo is able to classify traffic at line rate with nominal latency overhead, can scale to model sizes more than twice as large as state-of-the-art data plane ML classification systems, while achieving classification accuracy on-par with an offline traffic classifier.

  4. Jiang, C., Li, Z., Rao, S., & Tawarmalani, M. (2022). Flexile: Meeting bandwidth objectives almost always. Proceedings of the 18th International Conference on Emerging Networking Experiments and Technologies, 110–125.

    AbstractWide-area cloud provider networks must support the bandwidth requirements of network traffic despite failures. Existing traffic engineering (TE) schemes perform no better than an approach that optimally routes traffic for each failure scenario. We show that this results in sub-optimal routing decisions that hurt performance, and are potentially unfair to some traffic across scenarios. To tackle this, we develop Flexile, which exploits and discovers opportunities to improve network performance by prioritizing certain traffic in each failure state so that it can meet its bandwidth requirements. Flexile seeks to minimize a desired percentile of loss across all traffic flows, while modeling diverse needs of different traffic classes. To achieve this, Flexile consists of (i) an offline phase that identifies which failure states are critical for each flow; and (ii) an online phase, which on failure allocates bandwidth prioritizing critical flows for that failure state, while also judiciously allocating bandwidth to non-critical flows. For tractability, Flexile’s offline phase uses a decomposition algorithm aided with problem-specific accelerations. Evaluations using real topologies, and validated with emulation testbed experiments, show that Flexile outperforms state-of-the-art TE schemes including SWAN, SMORE, and Teavar in reducing flow loss at desired percentiles by 46% or more in the median case.

  5. Mathew, T. J., Tawarmalani, M., & Agrawal, R. (2022). Systematically identifying energy-efficient and attractive multicomponent distillation configurations. In Computer aided chemical engineering (Vol. 49, pp. 637–642). Elsevier.

    AbstractThousands of configurations exist for multicomponent distillation, making it laborious to use standard process simulators for identifying which among this plenitude are energy-efficient for a given separation. Shortcut models quickly screen the wide search space, but their development has been limited by various obstacles. In this work, we overcome three challenges: assumptions of constant relative volatilities and constant molar overflow, and utilizing heat integration. We incorporate our solutions into an optimization formulation and subsequently demonstrate its ability to identify energy-efficient and heat-integrated configurations on a case study. We also demonstrate how process intensification can be used to raise the value of the selected configuration.

  6. Chavez Velasco, J. A., Chen, Z., Gooty, R. T., Tawarmalani, M., & Agrawal, R. (2021). Energy-efficient membrane cascades for industrial separations. In Computer aided chemical engineering (Vol. 50, pp. 359–364). Elsevier.

    Abstracthe energy requirement for the separation of a given mixture via a multistage membrane cascade depends on the choice of the cascade and its operating conditions. Identifying the optimal cascade along with its optimal operating conditions is challenging, since it requires the solution of a nonconvex mathematical program. To address the challenge, we propose novel Mixed Integer Nonlinear Programs (MINLPs) that are formulated such that they can be solved using off-the-shelf global optimization solvers. We illustrate the practicality of our models with two case studies: (1) separation of p-xylene from o-xylene (2) recovery of natural gas liquid (NGL) from shale gas. Further, for NGL recovery, we determine the target selectivity and permeability that will enable membrane technology to outcompete the conventional demethanizer. These target values provide guidance for experimental groups that are developing new membrane materials for NGL recovery.

  7. Jiang, C., Rao, S., & Tawarmalani, M. (2020). PCF: Provably resilient flexible routing. Proceedings of the Annual Conference of the ACM Special Interest Group on Data Communication on the Applications, Technologies, Architectures, and Protocols for Computer Communication, 139–153.

    AbstractRecently, traffic engineering mechanisms have been developed that guarantee that a network (cloud provider WAN, or ISP) does not experience congestion under failures. In this paper, we show that existing congestion-free mechanisms, notably FFC, achieve performance far short of the network’s intrinsic capability. We propose PCF, a set of novel congestion-free mechanisms to bridge this gap. PCF achieves these goals by better modeling network structure, and by carefully enhancing the flexibility of network response while ensuring that the performance under failures can be tractably modeled. All of PCF’s schemes involve relatively light-weight operations on failures, and many of them can be realized using a local proportional routing scheme similar to FFC. We show PCF’s effectiveness through formal theoretical results, and empirical experiments over 21 Internet topologies. PCF’s schemes provably out-perform FFC, and in practice, can sustain higher throughput than FFC by a factor of 1.11X to 1.5X on average across the topologies, while providing a benefit of 2.6X in some cases.

  8. Chang, Y., Jiang, C., Chandra, A., Rao, S., & Tawarmalani, M. (2019). Lancet: Better network resilience by designing for pruned failure sets. Proceedings of the ACM on Measurement and Analysis of Computing Systems, 3, 1–26.

    AbstractRecently, researchers have started exploring the design of route protection schemes that ensure networks can sustain traffic demand without congestion under failures. Existing approaches focus on ensuring worst-case performance over simultaneous f-failure scenarios is acceptable. Unfortunately, even a single bad scenario may render the schemes unable to protect against any f-failure scenario. In this paper, we present Lancet, a system designed to handle most failures when not all can be tackled. Lancet comprises three components: (i) an algorithm to analyze which failure scenarios the network can intrinsically handle which provides a benchmark for any protection routing scheme, and guides the design of new schemes; (ii) an approach to efficiently design a protection schemes for more general failure sets than all f-failure scenarios; and (iii) techniques to determine which of combinatorially many scenarios to design for. Our evaluations with real topologies and validations on an emulation testbed show that Lancet outperforms a worst-case approach by protecting against many more scenarios, and can even match the scenarios handled by optimal network response.

  9. Gooty, R. T., Mobed, P., Tawarmalani, M., & Agrawal, R. (2018). Optimal multicomponent distillation column sequencing: Software and case studies. In Computer aided chemical engineering (Vol. 44, pp. 223–228). Elsevier.

    AbstractDistillation is one of the most widely used unit operation for separations in chemical and petrochemical industries. It is well-known that the number of distillation configurations available for the separation of an n-component mixture increases combinatorially with n. In this article, we describe a tool we have developed that screens through the entire search space to identify a handful of distillation configurations that are attractive for an application. Towards our goal, we formulate a novel Mixed Integer Nonlinear Program (MINLP) using Underwood’s method to estimate the vapour duty in each column. The MINLP formulation is integrated with DISTOPT: an easy-to-use in-house visualization software that takes feed properties as input, solves the optimization problem, and displays attractive configurations pictorially. The capabilities of the developed tool are illustrated with two case studies.

  10. Jiang, Z., Ramapriya, G. M., Tawarmalani, M., & Agrawal, R. (2018). Process intensification in multicomponent distillation. Chemical Engineering Transactions, 69, 841–846.

    AbstractProcess Intensification (PI) is an emerging concept in chemical engineering which describes the design innovations that lead to significant shrinkage in size and dramatic boost in efficiency in a process plant. Distillation, which is one of the most important separation technologies in the chemical industry, is therefore a crucial component in PI. Here, we discuss two aspects of PI in multicomponent distillation: 1) Performing simultaneous heat and mass integration among thermally coupled distillation columns to reduce the number of columns and heat duty requirement; and 2) Conducting any thermally coupled distillation in only a single column shell using a dividing wall column that is fully operable. Through examples, we show that synergistic use of both strategies leads to the design of compact, easy-to-operate, energy efficient and cost effective multicomponent distillation systems.

  11. Chang, Y., Rao, S., & Tawarmalani, M. (2017). Robust validation of network designs under uncertain demands and failures. 14th USENIX Symposium on Networked Systems Design and Implementation (NSDI 17), 347–362. URL

    AbstractA key challenge confronting wide-area network architects is validating that their network designs provide assurable performance in the face of variable traffic demands and failures. Validation is hard because of the exponential, and possibly non-enumerable, set of scenarios that must be considered. Current theoretical tools provide overly conservative bounds on network performance since to remain tractable, they do not adequately model the flexible routing strategies that networks employ in practice to adapt to failures and changing traffic demands. In this paper, we develop an optimization-theoretic framework to derive the worst-case network performance across scenarios of interest by modeling flexible routing adaptation strategies. We present an approach to tackling the resulting intractable problems, which can achieve tighter bounds on network performance than current techniques. While our framework is general, we focus on bounding worst-case link utilizations, and case studies involving topology design, and MPLS tunnels, chosen both for their practical importance and to illustrate key aspects of our framework. Evaluations over real network topologies and traffic data show the promise of the approach.

  12. Barik, A., Honorio, J., & Tawarmalani, M. (2017). Information theoretic limits for linear prediction with graph-structured sparsity. 2017 IEEE International Symposium on Information Theory (ISIT), 2348–2352.

    AbstractWe analyze the necessary number of samples for sparse vector recovery in a noisy linear prediction setup. This model includes problems such as linear regression and classification. We focus on structured graph models. In particular, we prove that sufficient number of samples for the weighted graph model proposed by Hegde and others is also necessary. We use the Fano’s inequality on well constructed ensembles as our main tool in establishing information theoretic lower bounds.

  13. Gençer, E., Tawarmalani, M., & Agrawal, R. (2015). Integrated solar thermal hydrogen and power coproduction process for continuous power supply and production of chemicals. In Computer aided chemical engineering (Vol. 37, pp. 2291–2296). Elsevier.

    AbstractIn this study, we introduce a novel solar thermal (ST) hydrogen and electricity coproduction process, which integrates a new solar water power (SWP) cycle and ST hydrogen production techniques. SWP cycle has a potential to generate electricity with efficiencies greater than 40% for solar heat collection temperatures above 1250 K. Higher solar heat collection temperatures enable the integration of SWP cycle with ST hydrogen production techniques to coproduce hydrogen and oxygen. When solar energy is unavailable, we propose a turbine based hydrogen water power cycle. For a 100 MW continuous power supply plant, the overall efficiency of the proposed integrated process is estimated to be greater than 34% in spite of the assumed 20% solar energy loss in the solar collector system. Furthermore, the coproduced hydrogen and oxygen can be used for various applications other than energy storage to create a sustainable economy.

  14. Shankaranarayanan, P., Sivakumar, A., Rao, S., & Tawarmalani, M. (2014). Performance sensitive replication in geo-distributed cloud datastores. 2014 44th Annual IEEE/IFIP International Conference on Dependable Systems and Networks, 240–251.

    AbstractModern web applications face stringent requirements along many dimensions including latency, scalability, and availability. In response, several geo-distributed cloud data stores have emerged in recent years. Customizing data stores to meet application SLAs is challenging given the scale of applications, and their diverse and dynamic workloads. In this paper, we tackle these challenges in the context of quorum-based systems (e.g. Amazon Dynamo, Cassandra), an important class of cloud storage systems. We present models that optimize percentiles of response time under normal operation and under a data-center (DC) failure. Our models consider factors like the geographic spread of users, DC locations, consistency requirements and inter-DC communication costs. We evaluate our models using real-world traces of three applications: Twitter, Wikipedia and Go Walla on a Cassandra cluster deployed in Amazon EC2. Our results confirm the importance and effectiveness of our models, and highlight the benefits of customizing replication in cloud datastores.

  15. Gençer, E., Mallapragada, D., Tawarmalani, M., & Agrawal, R. (2014). Synergistic biomass and natural gas conversion to liquid fuel with reduced CO2 emissions. In Computer aided chemical engineering (Vol. 34, pp. 525–530). Elsevier.

    AbstractTowards reducing the CO2 emissions associated with the transportation sector, we investigate the design of carbon and energy efficient processes for integrated biomass and natural gas (NG) conversion to liquid fuel. A process superstructure considering biomass conversion via gasification and Fischer-Tropsch (FT) synthesis or fast- hydropyrolsis and hydrodeoxygenation, and NG conversion via reforming followed by FT synthesis is established. Subsequently, a mixed integer nonlinear programming model (MINLP) is formulated to identify the process configurations that maximize the energy output as liquid fuel for different ratios of NG to biomass carbon feeds (delta-ng). For 1% <= delta-ng <= 150%, the optimal process configurations are capable of producing  5–14% more liquid fuel output than the combined fuel output of individual standalone processes converting the same amount of biomass and NG. This synergy originates from synthesizing additional liquid fuel by combining the residual biomass carbon with the excess hydrogen per carbon available from the NG feed. These integrated processes are also estimated to achieve up to 80% reductions in greenhouse gas (GHG) emissions relative to petroleum-based fuels.

  16. Ramapriya, G. M., Tawarmalani, M., & Agrawal, R. (2014). New, useful dividing wall columns for sustainable distillation. Book of Full Papers (Proceedings): 10th International Conference on Distillation & Absorption, 76–81.

  17. Nguyen, T. T., Tawarmalani, M., & Richard, J.-P. P. (2011). Convexification techniques for linear complementarity constraints. International Conference on Integer Programming and Combinatorial Optimization, 336–348.

    AbstractWe develop convexification techniques for linear programs with linear complementarity constraints (LPCC). In particular, we generalize the reformulation-linearization technique of Sherali and Adams to complementarity problems and discuss how it reduces to the standard technique for binary mixed-integer programs. Then, we consider a class of complementarity problems that appear in KKT systems and show that its convex hull is that of a binary mixed-integer program. For this class of problems, we study further the case where a single complementarity constraint is imposed and show that all nontrivial facet-defining inequalities can be obtained through a simple cancel-and-relax procedure. We use this result to identify special cases where McCormick inequalities suffice to describe the convex hull and other cases where these inequalities are not sufficient.

  18. Hajjat, M., Sun, X., Sung, Y.-W. E., Maltz, D., Rao, S., Sripanidkulchai, K., & Tawarmalani, M. (2010). Cloudward bound: Planning for beneficial migration of enterprise applications to the cloud. ACM SIGCOMM Computer Communication Review, 40(4), 243–254.

    AbstractIn this paper, we tackle challenges in migrating enterprise services into hybrid cloud-based deployments, where enterprise operations are partly hosted on-premise and partly in the cloud. Such hybrid architectures enable enterprises to benefit from cloud-based architectures, while honoring application performance requirements, and privacy restrictions on what services may be migrated to the cloud. We make several contributions. First, we highlight the complexity inherent in enterprise applications today in terms of their multi-tiered nature, large number of application components, and interdependencies. Second, we have developed a model to explore the benefits of a hybrid migration approach. Our model takes into account enterprise-specific constraints, cost savings, and increased transaction delays and wide-area communication costs that may result from the migration. Evaluations based on real enterprise applications and Azure-based cloud deployments show the benefits of a hybrid migration approach, and the importance of planning which components to migrate. Third, we shed insight on security policies associated with enterprise applications in data centers. We articulate the importance of ensuring assurable reconfiguration of security policies as enterprise applications are migrated to the cloud. We present algorithms to achieve this goal, and demonstrate their efficacy on realistic migration scenarios.

  19. Rahman, M. S., Kannan, K. N., & Tawarmalani, M. (2007). The countervailing incentive of restricted patch distribution: Economic and policy implications. WEIS. URL

    AbstractTraditionally, the government has been the sole entity to enforce anti-piracy measures. Of late, software vendors are attempting to thwart piracy of their products by providing patches only to legal users. By doing so, a vendor can vertically differentiate the legal copy from the pirated copy. It is not clear if the vendor’s differentiation strategy complements or substitutes the government’s effort with respect to social welfare. We study this issue by analyzing the impact of a monopolistic vendor’s action to restrict patches on both the vendor’s profit and the social welfare. Two key distinguishing features of our model are: (i) we endogenize the hacker activity and, therefore, the loss suffered by the users, and (ii) we also endogenize the quality of the patch developed by the vendor. Based on our analysis, we find that a monopolist does not always benefit from vertical differentiation. More specifically, when the government’s anti-piracy effort is intense and the cost of developing a good quality patch is high, the vendor does not benefit from vertical differentiation. Another interesting result of our analysis is that, by strategically utilizing the hacker’s activity, it is possible to improve social welfare relative to that when the patch is universally distributed.

  20. Xia, Q., Ersoy, O., Moskowitz, H., & Tawarmalani, M. (2007). Interactive clustering and classification. Proceedings of Conf. Artificial Neural Networks in Eng., (ANNIE’08), 463–470.

    Abstractn this paper, we propose a general framework referred to as interactive clustering and classification (ICC). It is designed to identify sub-groups of samples with different model structures. The framework features an interaction between classification and clustering process, allowing the clustering process to be partially driven by the classification process and is therefore, presumably, more informative. The method is tested rigorously on both synthetic datasets and real world problems. Experimental results demonstrate that ICC provided a good approximation of complex model structure by an aggregation of simple models while circumventing the issue of over-fitting.

  21. Tawarmalani, M., Kannan, K. N., & De, P. (2005). A mechanism for allocating objects in a network of symmetric caches. 15th Annual Workshop on Information Technolgies & Systems (WITS) Paper.

    AbstractIn this paper, we analyze object allocation in a network of caches that share web content to exploit network externality benefits. The analysis is presented for both centralized and decentralized scenarios, and is carried out using operations research and game-theoretic tools. The optimal allocation is found for each case, and cache incentives are aligned with the socially optimal welfare by devising appropriate pricing mechanisms.