Browsing by Author "Monis, L."

Now showing 1 - 7 of 7

DPDK-FQM: Framework for Queue Management Algorithms in DPDK
(Institute of Electrical and Electronics Engineers Inc., 2020) Pandey, A.; Bargaje, G.; Avinash; Krishnam, S.; Anand, T.; Monis, L.; Tahiliani, M.P.
The advantages of Network Function Virtualization (NFV) have attracted many use cases ranging from virtual Customer Premises Equipment (vCPE) to virtual Radio Access Network (vRAN) and virtual Evolved Packet Core (vEPC). Fast packet processing libraries such as Data Plane Development Kit (DPDK) are necessary to enable NFV. Currently, DPDK provides a framework for Quality of Service (QoS) which is used for queue management, traffic shaping and policing, but it lacks a general purpose queue management framework. In this paper, we propose DPDK-FQM, a framework to implement queue management algorithms in DPDK, run them and collect the desired statistics. Subsequently, we implement Proportional Integral controller Enhanced (PIE) and Controlled Delay (CoDel) queue management algorithms by using the proposed framework. We develop a new DPDK application to demonstrate the usage of APIs in DPDK-FQM, and verify the correctness of the framework and implementations of PIE and CoDel. Our experiments on a high speed network testbed show that PIE and CoDel exhibit their key characteristics by controlling the queue delay at a desired target, while fully utilizing the bottleneck bandwidth. Â© 2020 IEEE.
FQ-PIE Queue Discipline in the Linux Kernel: Design, Implementation and Challenges
(Institute of Electrical and Electronics Engineers Inc., 2019) Ramakrishnan, G.; Bhasi, M.; Saicharan, V.; Monis, L.; Patil, S.D.; Tahiliani, M.P.
Proportional Integral controller Enhanced (PIE) is an Active Queue Management (AQM) mechanism to address the bufferbloat problem. AQM mechanisms tackle bufferbloat by dropping or marking packets before the buffers fill up, but typically do not ensure fairness between responsive and unresponsive flows that share the same bottleneck link i.e., unresponsive flows can starve responsive flows when they co-exist. Recently, there has been an active interest in integrating flow protection mechanisms with AQM mechanisms to collectively tackle the problem of bufferbloat and fairness. There exist two such algorithms: Flow Queue Controlled Delay (FQ-CoDel) and Flow Queue Proportional Integral Controller Enhanced (FQ-PIE) that integrate flow protection with AQM mechanisms. Flow protection is achieved by dividing the incoming flows into separate queues and then applying CoDel/PIE algorithm on respective queues. Although FQ-CoDel is available in the mainline of Linux, there does not exist a model for FQ-PIE. In this paper, we discuss the design and implementation of FQ-PIE in the Linux kernel. We test and evaluate our proposed model of FQ-PIE in different scenarios by comparing the results obtained from it to those obtained for PIE and FQ-CoDel. Besides evaluating the fairness among responsive and unresponsive flows, we also evaluate the fairness among different types of responsive flows, such as when CUBIC TCP shares the same bottleneck link as TCP BBR. We also assess the benefits of integrating flow protection with AQM mechanisms in terms of reducing the latency for thin, latency sensitive flows when they coexist with thick, latency tolerant flows. Â© 2019 IEEE.
NeST: Network Stack Tester
(Association for Computing Machinery acmhelp@acm.org 2 Penn Plaza, Suite 701 New York NY 10121-0701, 2020) Rai, S.S.; Narayan, G.; Dhanasekhar, M.; Monis, L.; Tahiliani, M.P.
Linux network namespaces are a cost-effective and scalable alternative to physical systems for the design and experimental evaluation of network protocols. These evaluations are required for a practical understanding of how various networking algorithms would perform in the real world. However, manually setting up testbeds and obtaining results in the desired format using network namespaces can be quite cumbersome and error-prone. Although writing scripts could make these tasks easier, it becomes tedious and impractical if the network under consideration is large and complex. In this paper, we propose a python based package called NeST (Network Stack Tester) to perform tests for different congestion control algorithms and queue disciplines. It uses Linux network namespaces and provides APIs to create complex emulated networks, run tests and extract the statistics using iproute2 and netperf in a single python script. We validate the results obtained from NeST against those obtained from a physical testbed, and a virtual testbed setup manually by using network namespaces. The experiments with NeST are easy to reproduce because it is a wrapper around the existing tools and does not introduce new system dependencies. Â© 2020 ACM.
PowerDPDK: Software-Based Real-Time Power Measurement for DPDK Applications
(Institute of Electrical and Electronics Engineers Inc., 2020) Shah, M.; Yunus, M.; Vachhani, P.; Monis, L.; Tahiliani, M.P.; Talawar, B.
Data Plane Development Kit (DPDK) provides a set of libraries for fast packet processing that allow applications in the user space to directly interact with the NIC. Currently, DPDK provides a power management library that enables the applications to save power. However, it lacks features to effectively measure the power consumption of the system. In this paper we propose PowerDPDK, a software-based real-time library to measure the power consumption of DPDK applications. PowerDPDK leverages the Running Average Power Limit (RAPL) feature available on modern Intel processors to provide the power consumed by the CPU package and DRAM. We discuss the architecture of PowerDPDK and describe the process to incorporate it into DPDK applications. Subsequently, we use PowerDPDK to measure the power consumption of a few sample DPDK applications and a chain of Virtual Network Functions (VNFs) in OpenNetVM, a high-performance container-based platform for Network Function Virtualization (NFV). We show that a major share of the power consumed by DPDK is due to the use of Poll Mode Drivers (PMD), and hence, even a simple Layer 2 forwarding application consumes a large amount of power. Â© 2020 IEEE.
Programmable Data Plane for New IP using eXpress Data Path (XDP) in Linux
(IEEE Computer Society, 2022) Kataria, B.; Rohit, R.; Monis, L.; Tahiliani, M.P.; Makhijani, K.
This paper demonstrates a new dimension in packet programming and processing by leveraging New IP technology since applications are sensitive to different types of network constraints. For instance, emerging industry operations, manufacturing, and autonomics are limited by the stochastic quality of services and inflexible address structures. Instead, they require efficiency and deterministic networks. In this paper, we propose a programmable data plane for New IP packet processing and show how network headers could evolve in the future. We demonstrate the implementation of New IP stack to encompass three goals: (1) address customization: applications and routers can forward packets between hosts with different address formats. (2) design an end-to-end model to meet service delivery guarantees: routers implement various in-network New IP contracts as described by the applications. (3) Rapid experimentation of the New IP components. With New IP, developers can describe packet processing functionalities without depending on the specifics of the underlying hardware. Our implementation of New IP stack uses the existing toolsets and capabilities of the Linux platform, such as eXpress Data Path (XDP) and Traffic Control (TC) subsystem. It consists of an end-to-end solution with a new network stack on the host side and a corresponding packet processing and forwarding engine on the network. It is validated using Network Stack Tester (NeST), a lightweight tool built on Linux network namespaces. Â© 2022 IEEE.
Sobriety Testing Based on Thermal Infrared Images Using Convolutional Neural Networks
(2019) Kamath, A.K.; Karthik, A.T.; Monis, L.; Mulimani, M.; Koolagudi, S.G.
This paper proposes a method to test the sobriety of an individual using infrared images of the persons eyes, face, hand, and facial profile. The database we used consisted of images of forty different individuals. The process is broken down into two main stages. In the first stage, the data set was divided according to body part and each one was run through its own Convolutional Neural Network (CNN). We then tested the resulting network against a validation data set. The results obtained gave us an indication of which body parts were better suited for identifying signs of drunken state and sobriety. In the second stage, we took the weights of CNN giving best validation accuracy from the first stage. We then grouped the body parts according to the person they belong to. The body parts were fed together into a CNN using the weights obtained in the first stage. The result for each body part was passed to a simple back-propagation neural network (BPNN) to get final results. We tried to identify the most optimal configuration of neural networks for each stage of the process. The results we obtained showed that facial profile images tend to give very good indications of sobriety. The results also showed that combining the results of multiple body parts using a simple BPNN gives a higher accuracy than that of individual ones. � 2018 IEEE.
Sobriety Testing Based on Thermal Infrared Images Using Convolutional Neural Networks
(Institute of Electrical and Electronics Engineers Inc., 2018) Kamath, A.K.; Karthik, A.T.; Monis, L.; Mulimani, M.; Koolagudi, S.G.
This paper proposes a method to test the sobriety of an individual using infrared images of the persons eyes, face, hand, and facial profile. The database we used consisted of images of forty different individuals. The process is broken down into two main stages. In the first stage, the data set was divided according to body part and each one was run through its own Convolutional Neural Network (CNN). We then tested the resulting network against a validation data set. The results obtained gave us an indication of which body parts were better suited for identifying signs of drunken state and sobriety. In the second stage, we took the weights of CNN giving best validation accuracy from the first stage. We then grouped the body parts according to the person they belong to. The body parts were fed together into a CNN using the weights obtained in the first stage. The result for each body part was passed to a simple back-propagation neural network (BPNN) to get final results. We tried to identify the most optimal configuration of neural networks for each stage of the process. The results we obtained showed that facial profile images tend to give very good indications of sobriety. The results also showed that combining the results of multiple body parts using a simple BPNN gives a higher accuracy than that of individual ones. Â© 2018 IEEE.