• Awesome edge computing
• Simulators
• Frameworks
  • Engine
  • Networks
• Test (data, benchmark)
• Tools
• Applications
• Edge-AI frameworks
• Other awesome list

Awesome edge computing

Simulators

• CloudSim: CloudSim goal is to provide a generalized and extensible simulation framework that enables modeling, simulation, and experimentation of emerging Cloud computing infrastructures and application services, allowing its users to focus on specific system design issues that they want to investigate, without getting concerned about the low level details related to Cloud-based infrastructures and services.

• CloudSim Plus Automation: CloudSim Plus Automation is a Java 17+ command line tool based on CloudSim Plus (and some CloudReports classes) which is able to read specifications of CloudSim Plus simulation scenarios from a YAML file, a very human-readable data format. Simulation scenarios can be written inside a YAML file and CloudSim Plus Automation reads these simulation scenarios, creates and runs them on CloudSim Plus.

• CloudSim+ - Py4j gateway: py4j gateway for Cloud Sim Plus framework.

• CloudSimSDN: An SDN extension of CloudSim project to simulate SDN features in the context of a cloud data center. CloudSimSDN supports calculating power consumption by both hosts and switches. For instance, network-aware VM placement policies can be evaluated using CloudSimSDN.

• CloudSimPy: CloudSimPy is a data centric task schedule framework. It is bases on SimPy, a process-based discrete-event simulation framework based on standard Python. The scientific computing, deep learning, and machine learning ecology of the Python language is more complete than other programming languages. CloudSimPy works well with deep learning frameworks with Python support (such as TensorFlow, PyTorch) The combination helps to study resource management methods based on machine learning or deep learning.

• CFN: Computing First Networking that based on Named Data Networking (NDN).

• Cooja: Cooja Simulator is a network simulator specifically designed for Wireless Sensor Networks.

• CORE:The Common Open Research Emulator (CORE) is a tool for emulating networks on one or more machines. You can connect these emulated networks to live networks. CORE consists of a GUI for drawing topologies of lightweight virtual machines, and Python modules for scripting network emulation.

• DFaaS: A novel decentralized FaaS-based architecture designed to automatically and autonomously balance the traffic load across edge nodes belonging to federated Edge Computing ecosystems. Its implementation relies on an overlay peer-to-peer network and a distributed control algorithm that takes decisions on load redistribution.

• ECSNeT++: ECSNeT++ is a simulation toolkit for simulating the execution of Distributed Stream Processing applications on Edge anc Cloud Computing environments. ECSNeT++ is implemented using the OMNeT++ and the INET framework.

• EdgeCloudSim: EdgeCloudSim: An Environment for Performance Evaluation of Edge Computing Systems.

• EdgeSim:An open-source simulator of edge computing and caching. This simulator system contains three models: Content, Base Station and User.

• EmuFog: EmuFog helps to test fog computing applications more efficiently. Instead of actual deploying large network topologies with your application to test, EmuFog helps to generate networks that can be emulated easily with MaxiNet, a distributed version of the popular Mininet. This provides more realistic results than simulations and is cheaper and faster than real deployments. As an input EmuFog supports generated topologies from BRITE or measured real world topologies from Caida. In those networks EmuFog places fog nodes efficiently based on user defined constrains such as network latency thresholds or resource constraints. Applications for clients and fog nodes can be anything shipped in a Docker container.

• EnvisEdge: EnvisEdge allows users to simulate an edge computing environment to test their ideas and models before putting them in place on the edge. It takes care of all the complex stuff such as diversity across operating systems, computation power and communication mediums, allowing you to focus on the idea rather than the setup. EnvisEdge allows researchers, developers and data scientists to experiment and test their hypotheses, and produce production-ready code without having direct access to the edge devices. Creating a path for global research and growth in the domains of federated learning and edge computing.

• EPCSAC: The EPCSAC (Extensible Platform for Cloud Scheduling Algorithm Comparison) is an online open-source platform developed to help researchers compare scheduling algorithms to allocate tasks into virtual machines inside cloud infrastructures.

• FAAP-Simulator: Application allocation in the area of fog computing and targets fog networks in industrial environments. With the simulator, it is possible to generate, run and evaluate instances of the Fog Application Allocation Problem denoted as FAAP. The simulator is completely implemented in the scala programming language.

• faas-sim: Faas-sim is a powerful trace-driven simulation framework to simulate container-based function-as-a-service platforms. It can be used to develop, and evaluate the performance of operational strategies for such systems, like scheduling, autoscaling, load balancing, and others. faas-sim was developed at the Distributed Systems Group at TU Wien as part of a larger research effort surrounding serverless edge computing systems.

• Fogbed: Fogbed is a framework that extends the Mininet emulator to create fog testbeds in virtualized environments. Using a desktop approach, Fogbed enables the deployment of virtual fog nodes as Docker containers under different network configurations. The Fogbed API provides functionality to add, connect and remove containers dynamically from the network topology. These features allow for the emulation of real world cloud and fog infrastructures in which it is possible to start and stop compute instances at any point in time. Also, it is possible to change at runtime resource limitations for a container, such as CPU time and memory available.

• Fogify: Fogify is an emulation Framework easing the modeling, deployment and experimentation of fog testbeds. Fogify provides a toolset to: model complex fog topologies comprised of heterogeneous resources, network capabilities and QoS criteria; deploy the modelled configuration and services using popular containerized infrastructure-as-code descriptions to a cloud or local environment; experiment, measure and evaluate the deployment by injecting faults and adapting the configuration at runtime to test different "what-if" scenarios that reveal the limitations of a service before introduced to the public.

• FogNetSim++: FogNetSim++ extends OMNeT++, which is a well known framework for building network simulators, to model all these aspects. Moreover, it includes popular communication protocols for simulation, such as TCP, UDP, MQTT, and CoAP. Furthermore, FogNetSim++ models several other aspects, such as energy consumption, pricing, mobility, and handoff mechanisms.

• FogTorchPI: FogTorchΠ is an open source prototype, developed in Java, based on a model for Fog computing infrastructures and applications. 1), It takes into account non-functional parameters within the model (i.e., hardware, software, latency and bandwidth) to determine, compare and contrast different eligible deployments of a given application over a Fog infrastructure. 2), In the case of hardware capabilities, it considers CPU cores, RAM and storage available at a given node or required by a given software component. 3), Software capabilities are represented by a list of software names (operating system, programming languages, frameworks etc). 4), It considers latency, and both download and upload bandwidths as QoS attributes. Latency is measured in milliseconds (ms), while bandwidth is given in Megabits per second (Mbps). Note: Outputs of FogTorchPI can be the input of iFogSim.

• gem5: The gem5 simulator is a modular platform for computer-system architecture research, encompassing system-level architecture as well as processor microarchitecture.

• IoTSim-Edge: IoTSim captures the behavior of heterogeneous IoT and edge computing infrastructure and allows users to test their infrastructure and framework in an easy and configurable manner. IoTSim-Edge extends the capability of CloudSim to incorporate the different features of edge and IoT devices.

• IoTSim-Osmosis: IoTSim-Osmosis is a simulation framework that supports the testing and validation of osmotic computing applications. In particular, it enables a unified modelling and simulation of complex IoT applications over heterogeneous edge-cloud SDN-aware environments.DOI:10.1016/j.sysarc.2020.101956

• MARSSx86: MARSSx86 (MARSS for short) is a tool for cycle accurate full system simulation of the x86-64 architecture, specifically multicore implementations.

• MobFogSim: An extension which can support for Mobility of iFogSim. MobFogSim can model more generalised aspects related to device mobility and VM/container migration in the fog, e.g., user position and speed, connection handoff, migration policies and strategies, to name a few. More details can be found in MobFogSim: Simulation of Mobility and Migration for Fog Computing.

• Mini-NDN:Mini-NDN is a lightweight networking emulation tool that enables testing, experimentation, and research on the NDN platform. It was initially based on Mini-CCNx which was a fork of Mininet. Mini-NDN uses the NDN libraries, NFD, NLSR, and tools released by the NDN project to emulate an NDN network on a single system.

• MinNet: Mininet creates a realistic virtual network, running real kernel, switch and application code, on a single machine (VM, cloud or native), in seconds, with a single command:sudo mn. Mininet is also a great way to develop, share, and experiment with OpenFlow and Software-Defined Networking systems.

• Mininet-WiFi: Emulator for Software-Defined Wireless Networks.

• MobEmu: An opportunistic network emulator that can run a user-created routing or dissemination algorithm on a desired mobility trace or synthetic model.

• ndnSIM: The ndnSIM 2.0 is NS-3 module that implements Named Data Networking (NDN) communication model, the clean slate Internet design.

• NFaaS: Enabling to migrate functions closer to the user in ICN environment without a global view of the network.

• iFogSim: The iFogSim Toolkit for Modeling and Simulation of Resource Management Techniques in Internet of Things, Edge and Fog Computing Environments.

• lightMANO: A lightweight distributed service orchestrator designed for resource constrained multi-access edge computing environments.

• LEAF: A simulator for Large Energy-Aware Fog computing environments, based on CloudSim Plus. LEAF enables energy consumption modeling of distributed, heterogeneous, and resource-constrained infrastructure that executes complex application graphs.

• OMNeT++: OMNeT++ is a public-source, component-based, modular and open-architecture simulation environment with strong GUI support and an embeddable simulation kernel. Its primary application area is the simulation of communication networks, but it has been successfully used in other areas like the simulation of IT systems, queueing networks, hardware architectures and business processes as well.

• Packet Tracer: Packet Tracer is a cross-platform visual simulation tool designed by Cisco Systems that allows users to create network topologies and imitate modern computer networks. The software allows users to simulate the configuration of Cisco routers and switches using a simulated command line interface. Packet Tracer makes use of a drag and drop user interface, allowing users to add and remove simulated network devices as they see fit.

• PeerSim: PeerSim is a simulation environment for P2P protocols in java. Its features include easy configuration, extendability and high performance.

• pFogSim: pFogSim (/p/-fôg-/sɪm/) is a play off iFogSim; A simulator made to handle large-scale FOG networks with the HAFA Puddle Strategy to help evaluate the potential advantages/disadvantages within user-customizable scenarios.

• PureEdgeSim: a new simulator based on CloudSim Plus that is designed to simulate Cloud, Edge, and Mist computing environments. It allows to evaluate the performance of resources management strategies in terms of network usage, latency, resources utilization, energy consumption, etc. and enables the simulation of several scenarios such as the Internet of Things (IoT), connected vehicles/VANETs/MANET, Mist computing environments (peer-to peer networks such as mobile devices Cloud), and mobile Edge computing.

• RECAP-DES: RECAP Discrete Event Simulation Framework an extension for CloudSimPlus

• RECAP Simulator Framework: The RECAP Simulation Framework is an open source simulation framework designed to support experimentation of infrastructure with different description models for workload, user distribution, network topology, and (physical and virtual) resource placement. It is able to simulate different use cases including Virtual Network Functions (VNFs), Elasticsearch, smart cities, and virtual Content Delivery Networks (vCDN). The output of the RECAP Simulation Framework depends on the use case, but in general, it is a set of metrics predefined by the user such as bandwidth, resource consumption (CPU, memory, storage), network delay, energy consumption, active number of VMs, cache hits and cache misses.

• SatEdgeSim: A Toolkit for Modeling and Simulation of Performance Evaluation in Satellite Edge Computing Environments, which uses CloudSim Plus and PureEdgeSim as the underlying simulation framework.

• SUMO - ITS: "Simulation of Urban MObility", or "SUMO" for short, is an open source, microscopic, multi-modal traffic simulation. It allows to simulate how a given traffic demand which consists of single vehicles moves through a given road network. The simulation allows to address a large set of traffic management topics. It is purely microscopic: each vehicle is modelled explicitly, has an own route, and moves individually through the network. Simulations are deterministic by default but there are various options for introducing randomness.

• SimFaaS: This is a project done in PACS Lab aiming to develop a performance simulator for serverless computing platforms. Using this simulator, we can calculate Quality of Service (QoS) metrics like average response time, the average probability of cold start, average running servers (directly reflecting average cost), a histogram of different events, distribution of the number of servers throughout time, and many other characteristics.

• SimFlex: SimFlex is proceeding along two synergistic fronts: (1) Flexus is a powerful and flexible simulator framework that allows full-system simulation that relies heavily on well-defined component interface models to facilitate both model integration and compile-time simulator optimization. (2) SMARTS applies rigorous statistical sampling the­ory to reduce simulation turnaround by several orders of magnitude, while achieving high accuracy and confidence in estimates.

• SimMobility: SimMobility is an integrated mobility simulation platform that comprehensively simulates Future Mobility scenarios by integrating long, medium, and short-term travel behavior. Various mobility-sensitive behavioral models are integrated within the state-of-the-art scalable simulators to predict the impact of mobility demands on transportation networks, intelligent transportation services and vehicular emissions. The platform simulates the effects of a portfolio of technology, policy and investment options under alternative future scenarios. SimMobility encompasses the modeling of millions of agents, including pedestrians, drivers, phones, traffic lights, GPS, cars, buses, and trains, from second-by-second to year-by-year simulations and across countries.

• SimpleIoTSimulator: SimpleIoTSimulator® is an easy to use, IoT Sensor/device simulator that quickly creates test environments made up of thousands of sensors and gateways, all on just one computer. SimpleIoTSimulator supports many of the common IoT protocols. They include: MQTT, MQTT-SN, MQTT-Broker, CoAP, HTTP/s client, HTTP/s server, Modbus over TCP, BACnet/IP server, LoRa Gateway, LoRa Device.

• Simu5G: Simu5G is the evolution of the popular SimuLTE 4G network simulator that incorporates 5G New Radio access.

• SimuLTE: SimuLTE is an innovative simulation tool enabling complex system level performance-evaluation of LTE and LTE Advanced networks (3GPP Release 8 and beyond) for the OMNeT++ framework.

• SLEIPNIR: SLEIPNIR is a DAG scheduling simulator focused on mobile cloud/edge/iot infrastructures.

• Step-ONE: Simulated Testbed for Edge-Fog Processes based on the Opportunistic Network Environment simulator.

• SVL Simulator: An end-to-end autonomous vehicle simulation platform.

• The ONE: The ONE is a simulation environment that is capable of: 1) generating node movement using different movement models. 2) routing messages between nodes with various DTN routing algorithms and sender and receiver types. 3) visualizing both mobility and message passing in real time in its graphical user interface. ONE can import mobility data from real-world traces or other mobility generators. It can also produce a variety of reports from node movement to message passing and general statistics.

• Veins - ITS: Veins is an open source framework for running vehicular network simulations. It is based on two well-established simulators: OMNeT++, an event-based network simulator, and SUMO, a road traffic simulator. It extends these to offer a comprehensive suite of models for IVC simulation.

• Veins LTE - ITS: Veins LTE is a simulator for heterogeneous vehicular networks. It provides fine-grained simulation of vehicular networks based on IEEE 802.11p and TE.

• VirtFogSim: VirtFogSim is a MATLAB-supported software toolbox that allows the dynamic joint optimization and tracking of the energy and delay performance of Mobile-Fog-Cloud systems for the execution of applications described by general Directed Application Graphs (DAGs). Check the paper link for more details.

• YAFS(Yet Another Fog Simulator): YAFS is a simulator tool based on Python of architectures such as: Fog Computing ecosystems for several analysis regarding with the placement of resources, cost deployment, network design, ... IoT environments are the most evident fact of this type of architecture. The highlights points of YAFS are:

  • Dinamyc topology: entities and network links can be created or removed along the simulation.
  • Dinamyc creation of messages sources: sensors can generate messages from different point access along the simulation.
  • And for hence, the placement allocation algorithm and the orchestration algorithm, that are extended by the user, can run along the simulation.
  • The topology of the network is based on Complex Network theory. Thus, the algorithms can obtain more valuable indicators from topological features.
  • The results are stored in a raw format in a nosql database. The simpler the format, the easier it is to perform any type of statistics.

Frameworks

Engine

• Akraino Edge Stack: Akraino is a set of open infrastructures and application blueprints for the Edge, spanning a broad variety of use cases, including 5G, AI, Edge IaaS/PaaS, IoT, for both provider and enterprise edge domains. These Blueprints have been created by the Akraino community and focus exclusively on the edge in all of its different forms. What unites all of these blueprints is that they have been tested by the community and are ready for adoption as-is, or used as a starting point for customizing a new edge blueprint.

• Apache Edgent(incubating): Apache Edgent is a programming model and micro-kernel style runtime that can be embedded in gateways and small footprint edge devices enabling local, real-time, analytics on the continuous streams of data coming from equipment, vehicles, systems, appliances, devices and sensors of all kinds (for example, Raspberry Pis or smart phones). Working in conjunction with centralized analytic systems, Apache Edgent provides efficient and timely analytics across the whole IoT ecosystem: from the center to the edge.

• Apache OpenWhisk: Apache OpenWhisk is an open source, distributed Serverless platform that executes functions (fx) in response to events at any scale. OpenWhisk manages the infrastructure, servers and scaling using Docker containers so you can focus on building amazing and efficient applications.

• Baetyl: Baetyl is an open edge computing framework of Linux Foundation Edge that extends cloud computing, data and service seamlessly to edge devices. It can provide temporary offline, low-latency computing services include device connection, message routing, remote synchronization, function computing, video capture, AI inference, status reporting, configuration ota etc.

• Bochs: Bochs is a highly portable open source IA-32 (x86) PC emulator written in C++, that runs on most popular platforms. It includes emulation of the Intel x86 CPU, common I/O devices, and a custom BIOS. Bochs can be compiled to emulate many different x86 CPUs, from early 386 to the most recent x86-64 Intel and AMD processors which may even not reached the market yet.

• BOINC: BOINC lets you help cutting-edge science research using your computer. The BOINC app, running on your computer, downloads scientific computing jobs and runs them invisibly in the background.

• Distributed Strom: Distributed Storm is an extension of Apache Storm that supports the execution of distributed, adaptive, and QoS-aware scheduling algorithms. It introduces some key components on each worker node that enhance the system with adaptation capabilities, relying on a MAPE (Monitor, Analyze, Plan, and Execute) feedback loop periodically executed. Specifically, the newly introduced components are: the AdaptiveSchedulers, the QoSMonitors, and the WorkerMonitors. Logics of a Bolt can be regarded as an edge server processes the data.

• ENORM: The ENORM framework primarily addresses the deployment and load balancing challenges on individual edge nodes. ENORM operates in a three-tier environment, but a master controller does not control the edge nodes. Instead, it is assumed that they are visible to cloud servers that may want to make use of the edge. The framework allows for partitioning a cloud server and offloading it to edge nodes for improving the overall QoS of the application.

• EdgeGallery (Documents maintained in Chinese): EdgeGallery is a 5G edge computing open source project supported by device vendors, operators, vertical industry partners, etc. Its goal is to create a public platform for edge computing that meets the characteristics of 5G MEC "connectivity and computation", standardize open network capabilities (especially for 5G networks), and generalize the lifecycle-management process of MEC application development (testing, migration, and operation).

• EdgeMesh:EdgeMesh is a part of KubeEdge, and provides a simple network solution for the inter-communications between services at edge scenarios.

• EdgeX Foundry: EdgeX Foundry is a highly flexible and scalable open source software framework that facilitates interoperability between devices and applications at the IoT Edge.

• Folding@home: Folding@home (FAH or F@h) is a distributed computing project for simulating protein dynamics, including the process of protein folding and the movements of proteins implicated in a variety of diseases. It brings together citizen scientists who volunteer to run simulations of protein dynamics on their personal computers and scientific researchers. Insights from these data are helping scientists to better understand biology and providing new opportunities for developing therapeutics.

• FogAtlas: FogAtlas (evolution of the former Foggy platform) is a software framework aiming to manage a geographically distributed and decentralized Cloud Computing infrastructure that provides computational, storage and network services close to the data sources and the users, embracing the Fog Computing paradigm. FogAtlas is able to manage the so called Cloud-to-Thing Continuum offering service-aware workload placement and zero-touch deployment. It is an evolution of the well known paradigms of IaaS and PaaS adding the concept of “locality” to the traditional Cloud Computing model and easing the operations of a Fog Computing infrastructure.

• FogFlow: FogFlow is an IoT edge computing framework to automatically orchestrate dynamic data processing flows over cloud and edges driven by context, including system context on the available system resources from all layers, data context on the registered metadata of all available data entities, and also usage context on the expected QoS defined by users.

• k3OS: k3OS is purpose-built to simplify Kubernetes operations in low-resource computing environments. Installs fast. Boots faster. Managed through Kubernetes.

• K3s: Lightweight Kubernetes. Production ready, easy to install, half the memory, all in a binary less than 100 MB. Great for Edge, IoT, CI, Development, ARM, Embedding k8s, and Situations where a PhD in k8s clusterology is infeasible

• Krustlet: Krustlet: Kubernetes Kubelet in Rust for running WASM Krustlet acts as a Kubelet by listening on the event stream for new pods that the scheduler assigns to it based on specific Kubernetes tolerations. The default implementation of Krustlet listens for the architecture wasm32-wasi and schedules those workloads to run in a wasmtime-based runtime instead of a container runtime.

• KubeEdge:KubeEdge is an open source system extending native containerized application orchestration and device management to hosts at the Edge. It is built upon Kubernetes and provides core infrastructure support for networking, application deployment and metadata synchronization between cloud and edge. It also supports MQTT and allows developers to author custom logic and enable resource constrained device communication at the Edge. KubeEdge consists of a cloud part and an edge part.

• OCI: a framework that enables network operators with the ability to open up their edge facilities to Application Service Providers, by offering edge computing. As an example, consider a third-party Application Service Provider selling an IoT device (say, a home thermostat or security camera). Whenever that device shows up at the edge of the network, the Open Carrier Interface framework starts the Application Service Provider implemented edge software at the network operator’s Central Offices, which supports the IoT device with edge processing. Thus, even an early-stage Application Service Provider, who has few financial or physical resources, can offer the same level of edge support as giant companies, and therefore can compete with them on an even footing. read more.

• OpenStack++: is a framework developed by Carnegie Mellon University Pittsburgh for providing VM-based cloudlet platform on regular x86 computers for mobile application offloading. A set of new mobile computing applications that build upon OpenStack++ and leverage its support for cloudlets is also supported.

• OpenYurt: OpenYurt has been designed to meet various DevOps requirements against typical edge infrastructures. It provides the same user experience for managing the edge applications as if they were running in the cloud infrastructure. It addresses specific challenges for cloud-edge orchestration in Kubernetes such as unreliable or disconnected cloud-edge networking, edge node autonomy, edge device management, region-aware deployment and so on. OpenYurt preserves intact Kubernetes API compatibility, is vendor agnostic, and more importantly, is SIMPLE to use.

• SuperEdge: SuperEdge is an open source container management system for edge computing to manage compute resources and container applications in multiple edge regions. These resources and applications, in the current approach, are managed as one single Kubernetes cluster. A native Kubernetes cluster can be easily converted to a SuperEdge cluster.

• WebAssembly: Curated list of awesome things regarding WebAssembly (wasm) ecosystem.

• WebAssembly Micro Runtime: WebAssembly Micro Runtime (WAMR) is a standalone WebAssembly (WASM) runtime with a small footprint. It includes a few parts as below: 1, The "iwasm" VM core, supporting WebAssembly interpreter, ahead of time compilation (AoT) and Just-in-Time compilation (JIT). 2, The application framework and the supporting API's for the WASM applications. 3, The dynamic management of the WASM applications

• wasmCloud: wasmCloud is a distributed platform for writing portable business logic that can run anywhere from the edge to the cloud. Secure by default, wasmCloud aims to strip wasteful boilerplate from the developer experience and return joy to the act of building distributed applications.

• WasmEdge Runtime: WasmEdge (previously known as SSVM) is a lightweight, high-performance, and extensible WebAssembly runtime for cloud native, edge, and decentralized applications. It is the fastest Wasm VM today. WasmEdge is an official sandbox project hosted by the CNCF. Its use cases include serverless apps, embedded functions, microservices, smart contracts, and IoT devices.

• Wasmer: Wasmer is a fast and secure WebAssembly runtime that enables super lightweight containers to run anywhere: from Desktop to the Cloud, Edge and IoT devices.

• Wasmtime: Wasmtime is a Bytecode Alliance project that is a standalone wasm-only optimizing runtime for WebAssembly and WASI. It runs WebAssembly code outside of the Web, and can be used both as a command-line utility or as a library embedded in a larger application.

• WSO2-IoT Server: An extension of the popular open-source enterprise service-oriented integration platform WSO2 server that consists of certain IoT-related mechanisms, such as connecting a broad range of common IoT devices with the cloud using standard protocols such as MQTT and XMPP. Further, WSO2–IoT server includes the embedded Siddhi 3.0 component that allows the system to deploy real-time streaming processes in embedded devices. In other words, WSO2–IoT server provides the FEC computing capability to outer-edge devices.

Networks

• Awesome-pcaptools:A collection of tools developed by other researchers in the Computer Science area to process network traces.

• EdgeNet: edgeNet is a distributed edge cloud, in the family of PlanetLab, GENI, Canada’s SAVI infrastructure, Japan’s JGN-X, Germany’s G-Lab, and PlanetLab Europe. It is a modern distributed edge cloud, incorporating advances in Cloud technologies over the past few years.

• Mosquitto: Eclipse Mosquitto is an open source (EPL/EDL licensed) message broker that implements the MQTT protocol versions 5.0, 3.1.1 and 3.1. Mosquitto is lightweight and is suitable for use on all devices from low power single board computers to full servers. The MQTT protocol provides a lightweight method of carrying out messaging using a publish/subscribe model. This makes it suitable for Internet of Things messaging such as with low power sensors or mobile devices such as phones, embedded computers or microcontrollers. The Mosquitto project also provides a C library for implementing MQTT clients, and the very popular mosquitto_pub and mosquitto_sub command line MQTT clients.

• Naming Data Network Platform: NDN is an entirely new architecture, but one whose design principles are derived from the successes of today’s Internet, reflecting our understanding of the strengths and limitations of the current Internet architecture, and one that can be rolled out through incremental deployment over the current operational Internet.

• Node-RED: Node-RED is a programming tool for wiring together hardware devices, APIs and online services in new and interesting ways. It provides a browser-based editor that makes it easy to wire together flows using the wide range of nodes in the palette that can be deployed to its runtime in a single-click.

• Open vSwitch: is a production quality, multilayer virtual switch licensed under the open source Apache 2.0 license. It is designed to enable massive network automation through programmatic extension, while still supporting standard management interfaces and protocols (e.g. NetFlow, sFlow, IPFIX, RSPAN, CLI, LACP, 802.1ag). In addition, it is designed to support distribution across multiple physical servers similar to VMware's vNetwork distributed vswitch or Cisco's Nexus 1000V.

• POX: POX is a networking software platform written in Python. POX started life as an OpenFlow controller, but can now also function as an OpenFlow switch, and can be useful for writing networking software in general. An useful link of how to simulate a SDN can check the link: http://www.brianlinkletter.com/using-the-pox-sdn-controller/

• RICE: A unified approach to remote function invocation in ICN(Naming Data Network Platform) that exploits the attractive ICN properties of name-based routing, receiver-driven low and congestion control, low balance, and object-oriented security while presenting a natural programming model to the application developer.

• VerneMQ: VerneMQ is a high-performance, distributed MQTT message broker. It scales horizontally and vertically on commodity hardware to support a high number of concurrent publishers and consumers while maintaining low latency and fault tolerance. VerneMQ is the reliable message hub for your IoT platform or smart products.

• Wonder Shaper: Wonder Shaper is a script that allows the user to limit the bandwidth of one or more network adapters. It does so by using iproute's tc command, but greatly simplifies its operation.

Test (data, benchmark)

• AI Benchmark: The performance and comparison of all chipsets from Qualcomm, HiSilicon, Samsung, MediaTek and Unisoc that are providing hardware acceleration for AI inference.

• CloudSuite: CloudSuite is a benchmark suite for cloud services. The third release consists of eight applications that have been selected based on their popularity in today’s datacenters. The benchmarks are based on real-world software stacks and represent real-world setups.

• DeFog: DeFog, a first Fog benchmarking suite to: (i) alleviate the burden of Fog benchmarking by using a standard methodology, and (ii) facilitate the understanding of the target platform by collecting a catalogue of relevant metrics for a set of benchmarks.

• Edge AIBench: Edge AIBench is a benchmark suite for end-to-end edge computing including four typical application scenarios: ICU Patient Monitor, Surveillance Camera, Smart Home, and Autonomous Vehicle, which consider the complexity of all edge computing AI scenarios. In addition, Edge AIBench provides an end-to-end application benchmarking framework, including train, validate and inference stages.

• EUA Datasets: This repository maintains a set of EUA datasets which we collected from real-world data sources. The datasets are publicly released to facilitate research in Edge Computing. All the data is in Australia region which contains edge server locations and user location.

• MLPerf Inference Benchmark Suite: MLPerf Inference is a benchmark suite for measuring how fast systems can run models in a variety of deployment scenarios.

• networkX: NetworkX is a Python language package for exploration and analysis of networks and network algorithms. The core package provides data structures for representing many types of networks, or graphs, including simple graphs, directed graphs, and graphs with parallel edges and self-loops. The nodes in NetworkX graphs can be any (hashable) Python object and edges can contain arbitrary data; this flexibility makes NetworkX ideal for representing networks found in many different scientific fields.

• Neural Network Accelerator Comparison: A comparison of AI hardware accelerators among different platform.

• Rocketfuel: Traceroutes were sourced from 800 vantage points hosted by nearly 300 traceroute web servers. They started by mapping 10 ISPs, in Europe, Australia and the United States. In that process, they constructed a database of over 50 thousand IP addresses representing 45 thousand routers in 537 POPs connected by 80 thousand links.

• The CAIDA Anonymized Internet Traces 2015 Dataset: CAIDA's passive traces dataset contains traces collected from high-speed monitors on a commercial backbone link. The data collection started in April 2008 and ended in January 2019. These data are useful for research on the characteristics of Internet traffic, including application breakdown, security events, geographic and topological distribution, flow volume and duration.

• tinyMLPerf Deep Learning Benchmarks for Embedded Devices: The goal of TinyMLPerf is to provide a representative set of deep neural nets and benchmarking code to compare performance between embedded devices. Embedded devices include microcontrollers, DSPs, and tiny NN accelerators. These devices typically run at between 10MHz and 250MHz, and can perform inference using less then 50mW of power.

Tools

• ASSOLO: ASSOLO is a new active probing tool for estimating available bandwidth based on the concept of "self-induced congestion". ASSOLO features a new probing traffic profile called REACH (Reflected ExponentiAl Chirp), which tests a wide range of rates being more accurate in the center of the probing interval. Moreover, the tool runs inside a real-time operating system and uses some de-noising techniques to improve the measurement process.

• netem : netem provides Network Emulation functionality for testing protocols by emulating the properties of wide area networks. The current version emulates variable delay, loss, duplication and re-ordering.

• nmon: This systems administrator, tuner, benchmark tool gives you a huge amount of important performance information in one go. Many helpful tools for data analysis can be also found on section Data Analysis.

• GeoLite2: Mapping source IPv4 addresses to geo-graphical locations.

• Sigar: The Sigar API provides a portable interface for gathering system information such as:

  • System memory, swap, cpu, load average, uptime, logins
  • Per-process memory, cpu, credential info, state, arguments, environment, open files
  • File system detection and metrics
  • Network interface detection, configuration info and metrics
  • TCP and UDP connection tables Network route table

  This information is available in most operating systems, but each OS has their own way(s) providing it. SIGAR provides developers with API to access this information regardless of the underlying platform. one The core API is implemented in pure C with bindings currently implemented for Java, Perl, Ruby, Python, Erlang, PHP and C#.

Applications

• Edge Courier: An application for solving the whole-file-sync problem which needs high bandwidth in the cloud.

• Eman's Edge Computing System For AI Applications: Highly accurate AI applications, particularly for Computer Vision requires extensive memory and computational power. Eman's Edge Computing System orchestrates resource limited devices that requires using resource heavy AI algorithms so that a system can still achieve ideal system performances. This is done through Edge's monitoring modules that determines how to adjust the system so the system can operate efficiently based on its environments conditions.

Edge-AI frameworks

• AI Model Efficiency Toolkit (AIMET):AIMET is a library that provides advanced model quantization and compression techniques for trained neural network models. It provides features that have been proven to improve run-time performance of deep learning neural network models with lower compute and memory requirements and minimal impact to task accuracy.

• Apache TVM: Apache TVM is an open source machine learning compiler framework for CPUs, GPUs, and machine learning accelerators. It aims to enable machine learning engineers to optimize and run computations efficiently on any hardware backend.

• BerryNet: This project turns edge devices such as Raspberry Pi into an intelligent gateway with deep learning running on it. No internet connection is required, everything is done locally on the edge device itself. Further, multiple edge devices can create a distributed AIoT network.

• Bi-Real Net: Compared with the standard 1-bit CNNs, Bi-Real net utilizes a simple short-cut to significantly enhance the representational capability. Further, an advanced training algorithm is specifically designed for training 1-bit CNNs (including Bi-Real net), including a tighter approximation of the derivative of the sign function with respect the activation, the magnitude-aware gradient with respect to the weight, as well as a novel initialization.

• BMXNet: Apache MXNet is a deep learning framework designed for both efficiency and flexibility. It allows you to mix symbolic and imperative programming to maximize efficiency and productivity. At its core, MXNet contains a dynamic dependency scheduler that automatically parallelizes both symbolic and imperative operations on the fly. A graph optimization layer on top of that makes symbolic execution fast and memory efficient. MXNet is portable and lightweight, scalable to many GPUs and machines.

• BranchyNet: Fast inference via early exiting from deep neural networks. The architecture allows prediction results for a large portion of test samples to exit the network early via these branches when samples can already be inferred with high confidence.

• Caffe2:Caffe2 is a lightweight, modular, and scalable deep learning framework. Building on the original Caffe, Caffe2 is designed with expression, speed, and modularity in mind.

• CMSIS-NN: Collection of efficient neural network kernels developed to maximize the performance and minimize the memory footprint on Cortex-M processor cores.

• Compute Library: The Compute Library is a collection of low-level machine learning functions optimized for Arm® Cortex®-A and Arm® Mali™ GPUs architectures.

• Condensa: CONDENSA allows developers to design strategies for compressing DL models, resulting significant reduction in both the memory footprint and execution time. It uses a Bayesian optimization-based method to find compression hyperparameters automatically by exploiting an L-C optimizer to recover the accuracy loss during compression. They reported a 2.22 times runtime improvement over an uncompressed VGG-19 on the CIFAR-10 dataset with up to 65 times memory reduction. src from: 10.1145/3469029

• Core ML: Core ML optimizes on-device performance by leveraging the CPU, GPU, and Neural Engine while minimizing its memory footprint and power consumption. Running a model strictly on the user’s device removes any need for a network connection, which helps keep the user’s data private and your app responsive.

• daBNN: daBNN is an open-source fast inference framework developed by Zhang et al., which can implement Binary Neural Networks on ARM devices [161]. An upgraded bit-packing scheme and binary direct convolution have been used in this framework to shrink the cost of convolution and speed up inference. This framework is written in C++ and ARM assembly and has Java support for the Android package. This fast framework can be 6 times faster than BMXNet on Bi-Real Net. src from: 10.1145/3469029

• DDNN: distributed deep neural networks (DDNNs) over distributed computing hierarchies, consisting of the cloud, the edge (fog) and end devices. While being able to accommodate inference of a deep neural network (DNN) in the cloud, a DDNN also allows fast and localized inference using shallow portions of the neural network at the edge and end devices. Due to its distributed nature, DDNNs enhance data privacy and system fault tolerance for DNN applications. When supported by a scalable distributed computing hierarchy, a DDNN can scale up in neural network size and scale out in geographical span.

• DeepIoT: DeepIoT is a framework that shrinks a neural network into smaller dense matrices but keeps the performance of the algorithm almost the same. This framework finds the minimum number of filters and dimensions required by each layer and reduces the redundancy of that layer. src from: 10.1145/3469029

• DeepStack: DeepStack is an Open-Source AI API engine that serves pre-built models and custom models on multiple edge devices locally or on your private cloud.

• DeepThings: A framework for locally distributed and adaptive CNN inference in resource-constrained IoT edge clusters. DeepThings mainly consists of: 1). A Fused Tile Partitioning (FTP) method for dividing convolutional layers into independently distributable tasks. FTP fuses layers and partitions them vertically in a grid fashion, which largely reduces communication and task migration overhead. 2). A distributed work stealing runtime system for IoT clusters to adaptively distribute FTP partitions in dynamic application scenarios.

• Distiller: Distiller is an open-source Python package for neural network compression research. It provides a PyTorch environment for prototyping and analyzing compression algorithms, such as sparsity-inducing methods and low-precision arithmetic.

• FATE:FATE (Federated AI Technology Enabler) is an open-source project initiated by Webank's AI Department to provide a secure computing framework to support the federated AI ecosystem. It implements secure computation protocols based on homomorphic encryption and multi-party computation (MPC). It supports federated learning architectures and secure computation of various machine learning algorithms, including logistic regression, tree-based algorithms, deep learning and transfer learning.

• FedProx: FedProx aims to solve two key challenges that differentiate it from traditional distributed optimization: (1) significant variability in terms of the systems characteristics on each device in the network (systems heterogeneity), and (2) non-identically distributed data across the network (statistical heterogeneity).

• KitNET: KitNET is an online, unsupervised, and efficient anomaly detector. A Kitsune, in Japanese folklore, is a mythical fox-like creature that has a number of tails, can mimic different forms, and whose strength increases with experience. Similarly, Kit-NET has an ensemble of small neural networks (autoencoders), which are trained to mimic (reconstruct) network traffic patterns, and whose performance incrementally improves overtime.

• MACE: MACE (Mobile AI Compute Engine) is a deep learning inference framework optimized for mobile heterogeneous computing platforms. MACE provides tools and documents to help users to deploy deep learning models to mobile phones, tablets, personal computers and IoT devices.

• MegEngine: MegEngine is a fast, scalable and easy-to-use deep learning framework, with auto-differentiation.

• MindSpore Lite: MindSpore Lite is an ultra-fast, intelligent, and simplified AI engine that enables intelligent applications in all scenarios, provides E2E solutions for users, and helps users enable AI capabilities.

• ML Kit: ML Kit 19 is a mobile SDK framework introduced by Google. It uses Google's cloud vision APIs, mobile vision APIs, and TensorFlow Lite to perform tasks like text recognition, image labeling, and smart reply. Curukogluall et al. tested ML Kit APIs for image recognition, bar-code scanning, and text recognition on an Android device and reported that these APIs recognize different types of test objects such as tea cup, water glass, remote controller, and computer mouse successfully. src from: 10.1145/3469029

• MNN: MNN is a highly efficient and lightweight deep learning framework. It supports inference and training of deep learning models, and has industry leading performance for inference and training on-device. At present, MNN has been integrated in more than 20 apps of Alibaba Inc, such as Taobao, Tmall, Youku, Dingtalk, Xianyu and etc., covering more than 70 usage scenarios such as live broadcast, short video capture, search recommendation, product searching by image, interactive marketing, equity distribution, security risk control. In addition, MNN is also used on embedded devices, such as IoT.

• MQBench: MQBench is an open-source model quantization toolkit based on PyTorch fx. The envision of MQBench is to provide: 1) SOTA Algorithms. With MQBench, the hardware vendors and researchers can benefit from the latest research progress in academia. 2) Powerful Toolkits. With the toolkit, quantization node can be inserted to the original PyTorch module automatically with respect to the specific hardware. After training, the quantized model can be smoothly converted to the format that can inference on the real device.

• ncnn: ncnn is a high-performance neural network inference computing framework optimized for mobile platforms. ncnn is deeply considerate about deployment and uses on mobile phones from the beginning of design. ncnn does not have third party dependencies. it is cross-platform, and runs faster than all known open source frameworks on mobile phone cpu. Developers can easily deploy deep learning algorithm models to the mobile platform by using efficient ncnn implementation, create intelligent APPs, and bring the artificial intelligence to your fingertips. ncnn is currently being used in many Tencent applications, such as QQ, Qzone, WeChat, Pitu and so on.

• Neurosurgeon:A lightweight scheduler to automatically partition DNN computation between mobile devices and datacenters at the granularity of neural network layers.

• nn-Meter:nn-Meter is a novel and efficient system to accurately predict the inference latency of DNN models on diverse edge devices. The key idea is dividing a whole model inference into kernels, i.e., the execution units of fused operators on a device, and conduct kernel-level prediction.

• ns3-gym: OpenAI Gym is a toolkit for reinforcement learning (RL) widely used in research. The network simulator ns–3 is the de-facto standard for academic and industry studies in the areas of networking protocols and communication technologies. ns3-gym is a framework that integrates both OpenAI Gym and ns-3 in order to encourage usage of RL in networking research.

• NVIDIA TensorRT: NVIDIA® TensorRT™ is an SDK for high-performance deep learning inference. It includes a deep learning inference optimizer and runtime that delivers low latency and high throughput for deep learning inference applications.

• Once for All: Train One Network and Specialize it for Efficient Deployment.

• ONNX Runtime: ONNX Runtime is a cross-platform inference and training machine-learning accelerator. It can enable faster customer experiences and lower costs, supporting models from deep learning frameworks such as PyTorch and TensorFlow/Keras as well as classical machine learning libraries such as scikit-learn, LightGBM, XGBoost, etc. ONNX Runtime is compatible with different hardware, drivers, and operating systems, and provides optimal performance by leveraging hardware accelerators where applicable alongside graph optimizations and transforms.

• Paddle-Lite: Paddle Lite is an updated version of Paddle-Mobile, an open-open source deep learning framework designed to make it easy to perform inference on mobile, embeded, and IoT devices. It is compatible with PaddlePaddle and pre-trained models from other sources.

• PyTorch Mobile: The PyTorch Mobile runtime beta release allows you to seamlessly go from training a model to deploying it, while staying entirely within the PyTorch ecosystem. It provides an end-to-end workflow that simplifies the research to production environment for mobile devices. In addition, it paves the way for privacy-preserving features via federated learning techniques.

• SNPE: The Snapdragon Neural Processing Engine (SNPE) is a Qualcomm Snapdragon software accelerated runtime for the execution of deep neural networks. With SNPE, users can:

  • Execute an arbitrarily deep neural network
  • Execute the network on the SnapdragonTM CPU, the AdrenoTM GPU or the HexagonTM DSP.
  • Debug the network execution on x86 Ubuntu Linux
  • Convert Caffe, Caffe2, ONNXTM and TensorFlowTM models to a SNPE Deep Learning Container (DLC) file
  • Quantize DLC files to 8 bit fixed point for running on the Hexagon DSP
  • Debug and analyze the performance of the network with SNPE tools
  • Integrate a network into applications and other code via C++ or Java

• Tengine: Tengine is developed by OPEN AI LAB. This project meet the demand of fast and efficient deployment of deep learning neural network models on embedded devices. In order to achieve cross-platform deployment in many AIoT applications, this project is based on the original Tengine project using C language for reconstruction, and deep frame tailoring for the characteristics of limited embedded device resources. Also, it adopts a completely separated front-end/back-end design, which makes it possible to be transplanted and deployed onto CPU, GPU, NPU and other heterogeneous computing units rapidly, conveniently.

• TensorFlow Lite: TensorFlow Lite is an open-source deep learning framework to run TensorFlow models on-device. If you are new to TensorFlow Lite, we recommend that you first explore the pre-trained models and run the example apps below on a real device to see what TensorFlow Lite can do.

Other awesome list

• AI at the edge: A curated list of resources for embedded AI.
• Edge Computing & Internet Of Things: A qualitative compilation of production-ready edge computing projects with a focus on IoT based Data Engineering.