Developer-Led Landscape: Latency-Optimized Development
Could pushing apps and data closer to end users be a cloud opportunity worth $20 trillion?
With 5G, developers now vision the potential for low-latency applications.
When round-trip latency drops below 8ms, applications with real-time interactions emerge: remote surgery; remote-controlled vehicle control; jitter-less lag-less AR/VR conferencing; and just-in-time manufacturing driven by continuous supply and demand updates.
What are the developer-led landscape implications for the emergence of latency-optimized applications?
Previous Developer-Led Landscape Papers
Developer-Led Landscape: 2021 Developer Trends
Developer-Led Landscape: Cloud-Native Development
5 Architectures That Push Computation and Data (Closer) to the User
The lowest latency experiences are those where computation and data are co-located with the user.
Local-First / Offline-First Apps. Applications that operate without connectivity to networks. When connected to networks, can provide seamless experience and enhanced information, data, or services. User data is stored on the user’s device and owned by the user. The application may or may not synchronize data to other devices also controlled by the user. The application may or may not be installed / upgraded over a network. Local-first systems have no data “single source of truth” while offline-first systems do.
Technical Challenges: On-device persistence, user takes responsibility for management of user-owned data, offline / online transitioning, multi-device synchronization, app versioning and upgrades in partially connected or offline networks, reliable lightweight messaging for event transmissions, end-to-end encryption of communications and data on disk, cross-platform non-Internet peer-to-peer connectivity such as Bluetooth Low Energy or WiFi.
Vendor Conservatorship / Customer-Owned SAAS. A SAAS-styled architecture where the hosted systems are built by a vendor, but hosted by the paying customer. The customer can then operate local-first applications that access hosted data and services from their hosted, private implementation. The customer is the only party with visibility to the various end users operating different applications and, if there is any collected data about end users, that data is only accessible and controlled through the customer’s servers. The vendor is responsible for support, some remote operational management (if permitted by the customer), and to generate new versions. The vendor provides mechanisms to push and deploy new versions, which the customer updates through maintenance windows.
Technical Challenges: Synchronizing data structures between apps and SAAS, SAAS system versioning with disconnected apps, customer SAAS operations of a vendor-created system, remote-controlled vendor management of a customer-owned SAAS environment.
Geo-Fenced / Locale-Aware Network / Latency-Bounded. By reducing the number of “hops” a packet must travel to reach its destination, throughput and latency can be improved. This architecture enables traffic optimization, routing, and direct connections between clients and servers located within a geo-fence. For example, a user on a 5G network requests a video that is cached on a server nearby, the local server may be better suited to deliver the content than the nearest CDN. This architecture also enables applications to enable features only available for interacting with nearby devices, such as monitoring the Uber availability of cars within 5 miles of your current position. With fiber, an 8ms round trip with no intermediary processing is 1600km, so a system that guarantees <8ms round trip latency might be within a narrow geo-fence, maybe <100km; different latency boundedness may drive different geo-bounded variations driven by the physics of data transmission.
Technical Challenges: Geo-based routing and positioning, locale-specific feature implementations, time-space limit guarantees for round-trip processing.
Edge-Deployed SAAS. An application’s SAAS back-end that is hosted in an edge data center. This architecture can be tiered with different components of the SAAS back-end running on the edge or within centralized data centers. This includes the application server, databases, analytics, and observability components. In this architecture, a single “system” may replicate components to operate simultaneously on different edges to provide either redundancy or locale-specific performance and security enhancements. SAAS components that are running on the edge may be replicated across many edge nodes and need to incorporate distributed observability, data synchronization / replication, and coordination.
Technical Challenges: pushing system updates to all edge nodes, air gapped installations, data synchronization across edges, app deployment and versioning at the edge, integrating existing services such as AD, isolation of services operating on the edge, observability and coordination at scale (potentially 1 billion edges), data sovereignty across regulatory zones, access control for local groups, edge caching and preloading.
IOT-Specialized. Systems that blend different elements of the other architectures in a full stack offering that provides last-mile connectivity, edge processing, and centralized data center systems, typically tailored for the specialized qualities of a single type of IoT environment, such as commercial real estate, warehouse operations, or home automation.
Technical Challenges: Last mile device discovery and connectivity, event data aggregation / edge processing / aggregating, ML-on-the-edge, system monitoring for capacity and SLAs.
Local-Layered-Hybrid Architecture – Could it be the New Cloud Driving a $20 Trillion Transformation?
As with many things in life, blending the best concepts of different perspectives leads to something that is greater than the individual parts. Could that apply to architectures as well?
We are seeing that in the IoT space, the combination of private edge clusters, local networks, and local-first equipment is becoming more popular. And then these solutions become layered-hybrid by incorporating data center-hosted public clouds as a fallback strategy.
Local-first IOT devices -> call into an edge cluster on an as-needed basis, but does not require it to operate -> call back to private / public cloud services on an as-needed basis, but also does not need it to operate.
Global businesses experienced a $2 trillion dollar “digital transformation” makeover to create business systems that could deliver Web and mobile applications to end users with always-on centralized systems to service consumer’s data-hungry and service-needs.
The IoT industry has been going through a $100B “edge transformation” for the past 15 years to enable local hardware applications, partially-connected edge platforms, and corporate data center sinks and control planes.
But what happens if global businesses must go through a “local-edge-digital transformation” because their users demand minimal latency experiences? The popular layered architectures of IoT will need to be deployed for every business, reaching every locale, and applications will have to be “re-factored” to embrace offline-first and geo-fenced networks.
Holy moly – that will be a much more strenuous and challenging transformation. Forget $2 trillion, this could very well be the $20 trillion industry transformation.
Thank you, 5G.
On the other side of this transformation, the global distribution of computing resources will be diversified creating resiliency and redundancy for services that drive the betterment of humanity.
The Edge Lifecycle is Undefined – Limited Agility, Confusing Collaboration Models, and Multi-Variant Versioning Slows Edge Adoption and Creates Multi-Billion $$$ Opportunity for DevOps Vendors
Today’s developer lifecycle makes certain assumptions:
For a SAAS multi-tenant system, the versions running in production is one, or a limited fixed set.
The deviations between versions of software running on a client and a server system (and therefore the data + protocols exchanged between) is controlled and deterministic. This is further reduced with browser-only applications which re-load their client for each execution.
For SAAS systems, all users can be observed at all times.
An application’s surface area of attack vectors, whether SAAS or downloadable, is mathematically bounded and measurable, though admittedly it has practically been difficult to demonstrate.
Open source and vendor tools totaling more than $10B in ARR ranging from IDEs, version control, CI/CD, artifact repositories, observability / debugging, and threat monitoring are designed with team engagement models around these assumptions.
How do you perform continuous delivery with canary deployment when you cannot determine the number of offline clients you have?
How do you provide an audit check for data access when a billion distributed edge nodes takes 1 day to report and coalesce information?
What is the versioning and version control structure for a system that will have an unbounded number of versions of edge clients, edge nodes operating in the wild?
How do you debug systems that are eventually consistent, partially online, and must-not-crash?
Essentially, these architectures will require that every application have air gapped qualities. Few vendors offer air gap installation of their software, and those systems are typically in secure, no-network-isolation environments that ensure limited outside access. The way these systems are produced, versioned, upgraded, and managed differ from those that are network-connected. But in a latency-optimized world, applications may selectively depend upon their network and embrace air gapped, or partially connected network properties.
The world will need to rethink how their DevOps tools operate, and this will upend traditional vendors while creating opportunities for innovators.
CRDTs – Potentially Magical Data Structures For Data Convergence Across Time, Distance, and Objects?
As systems get pushed to the edge, the “same” data may simultaneously exist in multiple places: duplicates that reside on the edge away from the data source of truth, data pushed into clients even further away from the source of truth, or replicas across many clients in a local-first-no-source-of-truth configuration.
With the “same” data living simultaneously in different locations, modifications to that data must be converged with its replicas. Data convergence requires algorithms to address how changes will be organized, transmitted, and then merged into duplicate data structures. While simple to do in isolation, convergence becomes particularly challenging over long periods of time with large numbers of clients making numerous simultaneous changes.
There are two approaches to achieving eventual consistency. The first technique, popularized by Google Docs, was operational transforms (OT). In an OT system, each operation against a data set is sent to the replicas, and then each concurrent operation is transformed once received. OTs have been widely deployed, but after years of research, many OT algorithms have been found to not satisfy essential mathematical convergence properties, meaning that OT merging transformations are not always commutative.
CRDTs, introduced as a branch of computer science in 2011, are commutative. And thus, in a short decade, have become the focus of intense research and have been popularly deployed in a number of large scale systems. CRDTs, or conflict-free replicated data type, is a data structure that can be updated independently and concurrently without coordination between the replicas. CRDTs can either transmit their state changes … or … transmit their operations to their replicas. State-based CRDTs require more network bandwidth as local state must be transmitted to all other replicas and then merged by a function that is commutative, associative, and idempotent. Operation-based CRDTs require less network bandwidth, but require middleware to guarantee delivery of operations without duplication.
And thus, due to commutation desirability, CRDTs have received intense focus. Are they the magical data structure that will enable disparate systems to seamlessly converge, or will they create other headaches not yet felt? CRDTs do not allow invariants, which are things which cannot change. Envision a simple data structure that stores 1 item. If separate nodes simultaneously add an item, how can it converge?
Hmm… So maybe the world is getting closer, but more work needs to be done.
Latency-Optimized Development Platforms Await Watershed Application That Unlocks Multi-Decade, Trillion Dollar Opportunity
The technology a developer must learn to build an application that has a round-trip latency guarantee is significant.
Subsequently, the developer supply is limited, and the vendors providing edge platforms have modest adoption, growth and ARR.
Vendors and technology authors are awaiting for their edge Watershed moment: an application or workload emerges demanded by global populations that requires a latency guarantee <40ms. When an application emerges that becomes the must-have experience which can only work with a 8ms round trip latency, we’ll see the world transition towards new edge architectures, advanced 5G bandwidth and clients will be embraced, and developers will learn patterns for implementing local-first services.
Industrial IOT had been the original break through technology, with IIOT applications generating so much edge data that any IOT solution that doesn’t push computation and data analysis to the edge in real time isn’t viable. However, IIOT applications are specific to industry and often tailored to unique devices and protocols that cause their implementation to be walled gardens, with long cycle times, and rarely costing less than 7 figures. IIOT has not and will not be the Watershed moment.
What could be the Watershed App? I suppose it will need to be a consumer application potentially consumed by billions:
Competitive Gamers – gaining an edge through higher FPS or reactionary time offered through latency guarantees.
Autonomous Planes + Cars – ability for remote control and guidance of vehicles whether for fleet management or law enforcement safety.
Geo-fenced Commerce – Walk in, grab stuff, walk out with purchase transactions or violations captured before the person is out of reach.
The many vendors profiled in this article each have their own vision of what this might be. They all await patiently for it to emerge.
Technologies & Building Blocks
CRDTs: Automerge, Hypermerge, Gun, Yjs, Akka, Rust CRDT, Cloudstate, Complete List
P2P, edge network communication: WebRTC, IPFS / IPNS, IPFS Cluster, Dat, Relay(ers), BitTorrent, I2P, WiFi Aware, MQTT
Storing data on the local device: SQLite, Core Data, Firebase, Apple CloudKit, iOS Realm, Fission, ObjectBox
Multi-master replication databases: CouchDB, Cloudant, PouchDB, Hoodie
Push databases: RethinkDB
Offline-first caching: Rocicorp Replicache (client informs server which data it needs to sync), Attic Labs NomsDB (deprecated)
Distributed data: Akka Serverless (FaaS with databaseless distributed state through event sourcing), Cambria (schema translation between peers), Temporal.cloud, Swarm (decentralized storage), Nucypher (secrets management for decentralized apps), Gun (syncing decentralized graphs), Redwood (distributed database), Cloudflare Durable Objects (distributed objects with local state), Cloudflare Workers KV (distributed, eventually consistent key-value store)
Distributed observability: NS1 orb
End-to-end encryption: SSH (cryptographic network protocol for operating network services securely over an insecure network), TLS (cryptographic protocol to provide secure communication over a network), VPN (creates a private network connection across a public network connection).
Systems built with CRDTs: Azure Cosmos DB, Basho Riak, Weave Mesh (orphaned), SoundCloud Roshi (orphaned), Facebook OpenR (orphaned), Protocol Labs OrbitDB, Dappkit AvionDB, Macrometa, Akka Serverless, Concordant.io
Links To Stuff That Will Make You Smarter About Developing Latency-Sensitive Apps
Software feels slower
Measure software latency
CRDTs on Wikipedia
Server-to-server round-trip times between AWS and data centers
Hard parts of CRDTs
Peer to peer schema translation for distributed systems that change over time
Offline first application design
Google making web apps work offline
Offline app architecture
Toward confidential computing
John Mumm’s primer on CRDTs
John Mumm on convergence CRDTs
Data laced with history – causal trees and operational CRDTs
Neil Fraser on differential synchronization using fuzzy patching
Chas Emerick on distributed systems and the end of the API
Functional reactive programming
Convergence vs. consensus slides talk by Margin Kleppmann:
Coordination avoidance in database systems
Hypermerge: Automerge with DAT networking stack
IPFS local offline collaboration SIG
Ink & Switch - Seven Ideals For Local-First Software
Ink & Switch is an industrial research lab working on digital tools for creativity and productivity. They have multiple projects and experiments exploring the implications of local-first software.
No spinners: your work at your fingertips, ie – local access to data and instantaneous response to user input
Your work is not trapped on one device, ie – synchronization of your data to all your devices w/o configuration
The network is optional, ie – all necessary data, whether user’s or vendor’s, is stored locally on user’s device file system
Seamless collaboration with your colleagues, ie – collaboration on data, whether user’s or vendor’s, happens without central coordination
The long now, ie – continuous capabilities in the absence of vendor providing services because your data is stored locally
Security and privacy by default, ie – your device only saves your data and end-to-end encryption so sent data is encrypted
You retain ownership and control, ie – accessing your data is not limited by TOS, APIs and user takes data management responsibility
Rocicorp – 5 Principles For Edge Development
Rocicorp is a distributed software company building tools for creating distributed and offline systems.
Local-First: requires data ownership, ability to disconnect, take data with you, accepts P2P transmissions.
Offline-First: app built to run against local data, full functionality offline, conflict resolution after online sync must be designed into the app.
Real-time: local app receives updates as they happen on a server, conflict resolution from changing data must be designed into the app.
Multiplayer: objects are shared across different app instances run by different users, enabling users to interact simultaneously.
BYOBackend: enable the system of record to be controlled by the user, or hosted by the user within a regulatory zone they control.
Latency-Optimized Private Startup Vendor Profiles: $900M in VC Driving $400M ARR Generating $3B in Market Capitalization
The industry is incredibly lucky to have 30 startups that are driving innovations in and around the latency-optimized future. These vendors have raised $911M and collectively generate ~$400M in ARR, generating a fraction of the potential $20T value.
We are not covering public company edge platforms (often times offshoots of other products.
We also are not including CDNs, popularly considered an edge system for content acceleration, since they would skew the revenue numbers. However, if billions are already being spent on static content acceleration, that might be a fraction of what enterprises would spend to accelerate their dynamic systems. This might be a mistake because many CDNs are extending their platform with function execution platforms like AWS CloudFront Functions and Cloudflare Workers.
We did not profile MQTT vendors, a popular technology for bridging IoT device data to cloud systems, as many of these vendors provide MQTT capabilities as extensions to their classic messaging broker. MQTT vendors include Adafruit, EMQ Enterprise, Gurtam, GRID System, HiveMQ, IBM, Solace, Software AG, U-blox, and Bevywise Networks.
Public companies with edge, edge PAAS, or Industrial IOT include: Fastly, C3, Microsoft, PTC, Siemens, IBM, Software AG, Hitachi, SAP, AWS, GE Digital, Oracle, Google, VMware, Samsung, and Bosch.
CDNs include Netlify, Rackspace, Google, Akamai, Swarmify, Limelight, Microsoft, Amazon, Cloudflare, KeyCDN, SoftLayer, Cachefly, CDN77, and Imperva.
Increasingly, co-location and interconnect businesses like Equinix and Cyxtera (recently went public through a SPAC reverse merger) are also being called “edge” plays, providing co-location and bare metal solutions with non-5G interconnects that can offer 5ms round trip raw latency. We are not covering these edge building blocks in the discussion.
Edge SDK: Ditto.live
Founded in 2018. Raised $6.5M from Amity Ventures and True Ventures in 2020. CEO: Adam Fish, previously product at HBO, Realm, Roobiq.
Provider of a software development kit and application intended to help apps to sync with and even without connectivity. The company's software development kit directly communicates with other devices, without the need for a server and automatically manages the complexity of using multiple networks means of transport, like Bluetooth and WiFi, to find and connect to other devices and then synchronize any changes, enabling developers to build applications that sync with each other with and even without internet connectivity.
Edge SDK: fission.codes
Founded in 2018. Raised $550K from Lanebury Growth Capital and Outlier ventures in 2019. CEO: Boris Mann, previously doing a myriad of jobs in and around front-end development of applications.
It’s a development SDK that enables Web developers to create applications that run without servers and embedded within a browser. This includes mechanisms for cross-device synchronization, identity management and security, and device persistence. They are giving developers the ability to use front-end frameworks they love, but to get a local-first experience if a device is not connected to a network. A presentation on how their architecture works.
Edge SDK: Rocicorp
Founded in 2020. Bootstrapped with public statement to avoid VC (my feelings are hurt!). CEO: Aaron Boodman, Erik Arvidsson, and Fritz Schneider all previously from Google and Attic Labs.
Rocicorp is extending the original work done on the Noms database to create a series of tool kits for software developers to enable offline-first, concurrent multi-user, and bring your own back-end systems development. Their first effort is the open source Replicache, an offline-first caching system where the client informs the server of which data it wishes to cache and keep synchronized. This is quite different from a replicated database where the amount of data to be replicated is unwieldy along with increasing the chance of collisions from overlapping changes made on the same data across clients.
Edge SDK: Teraki
Founded in 2014. Raised $16.8M from Horizons Ventures, Innogy Innovation Hub, State Auto Labs, Paladin Capital Group, American Family Ventures, and MobilityFund. CEO: Daniel Richart, previously a researcher in quantum cryptography and computation.
Teraki has a suite of edge SDKs and libraries for embedded processing of time series, video codecs, video object classification, and 3D point rendering. Their SDKs are small, lightweight, and performant to work in real time applications on low-powered devices ideal for automotive and drone usage.
Edge DB: HarperDB
Founded in 2017. Raised $8.15M from Break Trail Ventures and Service Provider Capital. CEO: Stephen Goldberg, previously working in infrastructure and IoT with Phizzle, Kloudroot, and Red Hat.
Edge DB: Skyfoundry
Founded in 2009. Raised an undisclosed amount from Kohl’s and Battery. CEO: Brian Frank, previously Tridium and author of Fantom programming language.
Skyfoundry has designed a lightweight analytics system that can run in the cloud or on the edge optimized for data coming from IoT devices. Their software helps domain experts to capture their knowledge in rules and algorithms that automatically run against collected data, it includes features to organize and manage real-time and historical time-series data, it also supports machine learning and artificial intelligence pattern recognition as well as comprehensive data visualization, enabling clients to predict the growth of their business and visualize their data using their suitable browser. It includes pre-built applications for visualizing data including spark trending, KPIs, energy analysis, weather analysis, and historian tracking. Based upon the open source Project Haystack, an open source initiative supported by Siemens to create technologies for modeling IOT data.
Edge DB: ObjectBox
Founded in 2015. Raised $6.4M from Vito Ventures, Cavalry Ventures, and Techstars. CEO: Vivien Dollinger, previously Google, Koch Media, and Travian Games.
ObjectBox has built a lightweight database designed to be embedded on edge devices. It promises nearly 10x write performance vs. alternatives along with embedded synchronization to other ObjectBox databases. The database has been adapted with variants for working with IOT sensor data and time series information. Promises <1MB installation footprint and solid support for a wide range of programming languages. Surprised that this product doesn’t get way more love.
Edge PaaS: Mimik
Founded in 2009. Funding undisclosed. 39 employees. CEO: Fay Arjomandi, longtime telco executive with Vodafone, finsphere, Disternet, Mobidia, L3 Technology. Cathie Wood of ARK joined their board in 2021.
Cloud computing platform that decentralizes the cloud. Mimik creates clusters by physically discovering and establishing peer-to-peer connections between one another, enabling application hosting by extending the cloud to edge devices. Positions their edgeEngine product as a serverless offering that can be installed on edge devices including consumer devices and then app devs can build applications that access compute and data locally. I am thinking of Mimik as a micro Service Mesh that operates on mobile devices and on your cloud nodes, transparently handling routing and security concerns for API calls in between apps, other mobile devices, and the cloud.
Edge PaaS: Actyx
Founded in 2015. Raised an undisclosed amount of VC from Paua Ventures in 2017. CEO: Oliver Stollmann, previously a researcher at PARC in industrial automation.
Provides a developer platform for factory-based applications. It contains a twin programming model as defined by events and event reactions, local computing environments around physical assets, special cooperation among twin assets, and twin synchronization. Actyx has a programming model that focuses on process orchestration and abstracts IT. Provides CLI and management consoles for managing factory nodes and the apps running on the edge. ActyxOS layer requires Docker, Android, Windows, Linux, or MacOS to run. ActyxOS requires TypeScript, C#, or Rust to create applications. Guarantees eventually consistent business logic in a completely decentralized environment without central coordination. They are working to establish a concept of a “Local Twin”, an active and locally-evaluated model of an asset or workflow, a contrast to the increasingly popular “Digital Twin”, which is a passive representation of an asset or workflow used for simulations.
Edge PaaS: Lightbend (A DTC Portfolio Company!!)
Founded in 2011. CEO: Mark Brewer, previously Spring and VMware.
Akka Serverless is a stateful serverless platform that doesn’t require a database and provides a functions-as-a-service (FaaS) model using different programming languages. The biggest bottleneck to scaling a system that can be resilient under duress is state. Cloud systems that are based upon databases increase access latency and move the point of failure into the database tier. Planet-scale databases help alleviate the problem, but still have concentrated (locale-based) data access, usually pinned to a single region or location, nor do they have mechanisms for pushing actions and state to the edge or into local-first / offline-first clients. The best FaaS model for cloud development is one that treats events, functions and state each as first class constructs. Akka’s current design evolves around "state models" where state management is abstracted away and managed on behalf of the function (an entity as backed by an actor), and the runtime makes sure that an entity always sees and works with the latest data, on an as-needed basis. The state models are implemented transparently with Reactive Architecture concepts: Event Sourcing, CRDTs, and a simple CRUD-like (snapshot storage on function key). Akka Serverless is built with Akka (actor-based reactive programming model), Cloudstate (for addressing the abstractions of functions into state objects for persistence), gRPC (for asynchronous, decoupled communications), GraalVM, and KNative (for running FaaS systems on top of Kubernetes.
Edge PaaS: Macrometa
Founded in 2017. Raised $7.85 from DNX Ventures, Benhamour Global Ventures, Sway Ventures, Partch, Shasta Ventures, VU Venture Partners, Fusion Fund, and Velar Capital. CEO: Chetan Venkatesh, previously CEO of Atlantis Computing.
Macrometa provides a stateful serverless platform that combines a noSQL database, pub/sub messaging, event processing and computing platform to enable building geo-distributed applications. They focus on how to enable an application to be brought closer to end users. It is a geo-distributed, real-time, coordination-free materialized views engine. They have implemented a custom data model, materialized views, and multi-model query engine on top of a convergence engine that uses a CRDT operation log over streams. Data can be saved in custom key-value, document, graph, and stream structures.
Edge PaaS: Concordant
Founded in 2019. Raised an undisclosed amount. CEO: Mark Shapiro, previously the inventor of CRDTs and longtime Microsoft Research alum, and Peter Lash, previously Beacon Wireless.
Concordant is building a software layer that uses CRDTs and CRDT extensions to push and pull data to the edge. They are starting off as a software delivery but going to deliver a BaaS platform for app developers to create edge applications. Applications will be able to run on the edge or move to different locations with built-in collaboration. They embrace a local-first principle where all processing starts on the edge and only moves to the cloud if necessary.
Edge PaaS: The Deno Company
Founded in 2020. Raised $4.9M from Four Rivers Ventures, Rauch Capital, Long Journey Ventures, Mozilla Corporation, Shasta Ventures, and Ben Noordhuis. CEO: Ryan Dahl, Bert Belder previously the originators of Node.js.
On the surface, the Deno Company could be construed to be the commercial entity supporting the MIT-licensed Deno application server, a Node-like lightweight, parallel runtime with support for TypeScript and authored in Rust. However, the commercial entity has launched Deno Deploy, an edge compute platform that executes lightweight Deno processes on distributed edge nodes. Deno can be dynamically repackaged to exclude unnecessary subsystems, making it a type of chameleon ideal for running with a light footprint on edge nodes. The deploy solution lets developers author stateless functions locally in most any language which can be compiled into WebAssembly and then push to Deno’s distributed edge compute platform.
Edge PaaS: Stackpath
Founded 2015. Raised $606M from Sweetwater Private Equity, SymbaSafety, EMCJuniper Networks, Cox Communications, WhiteHorse Finance, NewStar Financial, Goldman Sachs, and Antares Capital. CEO: Christopher Turco, previously Abry Partners, SiteLock, Motricity, and Sungard.
Stackpath is an edge computing platform that provides infrastructure for edge processing and content caching. They operate edge locations that are co-located in major markets connected to a private network backbone.
Edge PaaS: AlefEdge
Founded in 2013. Raised $40M from Select Venture Partners, Tata Capital, AlphaPrime Ventures, Redwood Partners, and angels. CEO: Ganesh Sundaram, previously Alcatel-Lucent and Bell Labs.
AlefEdge wins the award for having the most confusing description of edge services out of this category. They provide an edge execution environment that allows application developers to push content, data, and video to 5G edge points so that applications written with Alef APIs can have low latency, high performance access to those services. They are partnering with 5G companies to create 100s of endpoints where application data can be pushed to. AlefEdge also provides 5G infrastructure which enables operators to stand up private 5G networks. Their service is still in beta.
Edge Acceleration: Outsmartly
Founded in 2020. Raised an undisclosed amount from Mango Capital. CEO: Shalom Volchok, previously Digital Optimization Group.
Outsmartly has built a CDN for dynamic components of a Web site. Where most CDNs cache static, unchanging content, Outsmartly provides hooks within a front-end’s dynamic processing components which allow its edge nodes to cache the results of dynamic computations for serving similar requests with a lower latency. The performance improvements of Jamstack applications is significant, though to capture the gains application developers need to modify how their front end systems are coded.
Edge Infrastructure: Mutable
Founded in 2012. Raised $1.61M by Lunar Ventures and Fly Ventures. Part of the 5G Open Innovation Lab. CEO: Antonio Pellegrino, previously CEO of LSQ (Heroku-like platform).
Mutable provides an edge OS to enable operators to run a public edge cloud. Mutable is a runtime platform for deployment onto edge servers to create micro-datacenters. They offer multi-tenant cloud isolation, runs their software through containers, and no-downtime deployments through blue-green strategies. Built using NixOS, wireguard secure transport, and a multi-cluster Kubernetes. You can access public edge clouds powered by Mutable’s software layer through partnerships they have with telecom and cable operators. They enable VMs, containers, and “serverless scripts” to run within their edge locations offering basic computation closer to the end user. They also have a low code approach to defining rules for redirects, header management, and URL signing that can execute on the edge.
Edge Infrastructure: Saguna
Founded in 2008. Raised $16M from CEIIF, Mogan Stanley, Kleiner Perkins, IDG Capital, Softbank Ventures, Akamai Technologies, iVentures, and Xenia Venture Capital. CEO: Ido Gur, previously Israeli Air Force.
Saguna makes an edge networking, routing, and compute platform for applications to execute on the edge. Saguna pushes key network services including IPSec, routing, switching and load balancing to execute inline within edge devices. This is layered with a compute environment that can execute VMs and containers that can be pushed and managed out on the edge through a set of APIs.
Edge Infrastructure: Edgeworx
Founded in 2017. Raised an undisclosed amount from Plug and Play Tech Center, Samsung NEXT Ventures, CloudScale Capital Partners, and Sequoia Capital. CEO: Kilton Hopkins, previously Northeastern University, iotracks, ICON Technology Corporation, JPMorgan Chase.
Edgeworx originally began as the commercial support around the open source Eclipse ioFog edge platform. ioFog provides a compute and execution platform for application services that runs on edge devices, creating a microservices environment. The runtime environment is extended with an edge network which provides a broker, secure traffic, routing, and remote control of services. Edgeworx has evolved to also produce Darcy, a converged hardware+software device which is running Eclipse ioFog with a built-in camera, thermal detection, and audio. The device is plug-and-play ready, but can be extended with custom services that leverage the analytics happening on the device. The Darcy device can be combined with Darcy Cloud PaaS which provides system monitoring, fleet management, remote access, cloud logging, self-healing, over the air updates, and self service SaaS portal.
Edge Infrastructure: Hivecell
Founded in 2015. Raised $8M from Investment Capital Ukraine and Nvidia. CEO: Jeffrey Ricker, previously Point72 Asset Management, Amazon, and State Street.
Hivecell is creating a fully managed hardware device that packages compute, storage, and networking into a plug-n-play, fully meshed environment for edge deployment. The embedded edge device is Kubernetes and Kafka-ready, so containerized and event-driven applications are suitable for deployment by customers. Hivecell provides remote administration and managed services for the edge devices.
Edge Infrastructure: Taubyte
Founded in 2020. Raised an undisclosed amount from an accelerator in 2020. CEO: Samy Fodil, previously Facebook and Sage Hero.
Taubyte has created a software edge stack that provides fully distributed key-value store, Mongo API-compatible database, ledger, and storage. The data components are replicated in between different Taubyte edge nodes. Developers can write procedural functions that execute on the edge system and make use of asynchronous message brokers. They are targeting this edge system for industrial, energy, and security field solutions.
Edge Infrastructure: Kubermatic
Founded in 2016. Raised an undisclosed amount. CEO: Sebastian Scheele, previously SAP.
Kubermatic packages a Kubernetes distribution that is optimized for edge execution. The platform can be installed in air gapped environments, execute within security zones, and deliver a cloud native experience without Internet connectivity. They also run on bare metal using KubeVirt virtual machines and also provide Kubernetes large cluster fleet management orchestration 1000s of clusters running in different locations.
Edge Encryption: Ockam
Founded in 2017. Raised $5.1M from Future Ventures, Okta Ventures, Samsun NEXT Ventures, Core Ventures Group, and SGH Capital. CEO: Matthew Gregory, previously Microsoft, Heroku, and Weather Underground.
Ockam is building a developer-friendly toolkit for creating applications which have end-to-end encryption and can guarantee various levels of trust to their end users. The end-to-end encryption can be between end user applications and server systems, or directly between two applications with an open channel over an open network. The credentials created with Ockam are verifiable and can represent unique devices. Ockam also provides a hosted service that will act as an intermediary of trusted messages between Ockam-authenticated connections.
Edge IAAS: Section
Founded in 2012. Raised $11.65M from Foundry Group, Next Frontier Capital, Lava Walk Ventures, and Secure Octane. CEO: Stewart McGrath, previously a CIO at GraysOnline and accounting at E&Y.
Developer of an edge compute platform to help developers to deploy microservice edge workloads using DevOps-centric tooling. They use ML algorithms to automate infrastructure provisioning, workload orchestration, infra scaling, and traffic routing --- combined through a location optimizer, traffic director, and endpoint controller. They distribute Kubernetes clusters across different infrastructure providers including AWS, Google, Azure, Digital Ocean, Packet, and Rackspace. They allow engineers to choose the edge modules that best fit their application, provides the flexibility to run workloads anywhere along the edge compute continuum, and modern DevOps principles with workflows that support the full development lifecycle, enabling developers to have control over their content delivery.
Edge IAAS: Fly.io (A DTC Portfolio Company!!)
Founded in 2016. Raised $3.65M from various investors as part of a YC round in spring 2020. CEO: Kurt Mackey, previously CEO of Compose and technology at Conde Nast and Ars Technica.
Fly provides a hosting environment that let’s developers deploy application servers and databases close to end users. Fly also provides a specialized routing layer that optimizes end user traffic to reach an endpoint with the lowest latency. Applications are deployed as Docker containers and also provides a specialized database tier that moves PostgreSQL databases to run on the edge, creating read replicas across each edge node with a clever technique for ensuring writes are routed to the appropriate master. As such, Fly offers a type of IAAS / CDN deployment environment for edge hosting, where the various micro data centers they run constitute the edge. Application teams leveraging Fly can get consistent <50ms round trip latency with database writes globally.
Edge IAAS: Seaplane.io
Founded in 2020. Raised $2M from Sequoia. Co-founders include Catherine Schikkerling, Denzil Dunn, Niall Dalton, and Suneil Mishra.
They produce a container IAAS that runs a single global zones that computes where and when a container should be activated for execution. They describe it as a single global cluster. Based upon a customer’s resource expectations, Seaplane adjusts physical resources around the globe. There is a way to add other infrastructure services including messaging, object storage, and geo-distributed databases.
Edge IAAS: Zededa
Founded 2016. Raised $28.5M Energize Ventures, Lux Capital Management, Rockwell Automation, Juniper Networks, EDF North America Ventures, Almaz Capital, and HBAM. CEO: Said Ouissal, previously Violin Memory, Juniper Networks, and Ericsson.
Zededa provides an edge infrastructure solution that promises to deploy any kind of cloud native app (containers or VMs) on any kind of edge hardware. Their kernel, called EVE-OS, was donated to the Linux Foundation’s LF Edge organization and focuses on maintaining software integrity by hardware root of trust. Zededa provides management apps for fleet management of devices and applications, open APIs for accessing the orchestration system, and a distributed firewall which can create overlay networks.
Edge Data: Alluxio
Founded in 2015. Raised $23M from Andreessen Horowitz, Volcanics Venture, and Seven Seas Partners. CEO: Haoyuan Li, previously Tachyon Nexus and Berkeley Computer Science.
Alluxio has created a data orchestration runtime which dynamically moves data to be closer to compute. The data orchestration layer creates a stronger separation between data and storage, eliminates data duplication, and can enable data to be moved from legacy stores into modern systems like object stores. Interestingly, while most vendors in the edge space are working to move data closer to the end user, Alluxio’s data orchestration provides more value by moving big data workloads across data centers or in bursting scenarios to optimizes the co-location of compute and data in big data processing scenarios. Edge systems can benefit from Alluxio as a tool to reduce the number of round trip queries, reduce the cost of interactive querying over remote networks, and to reduce the costs of sharing large datasets across queries.
Edge API: Droplit
Founded in 2014. Raised $520K from Mosley Ventures and Florida Funders. CEOs are Bryan Jenks previously software engineers at various locations and Stephen Westerfield previously an analyst at Morgan Stanley.
Droplit provides an IOT integration platform. They provide access and control for dozens of connected products through a single API. They target ISVs who need to build an ecosystem of integrations as part of their application. For example, if you were building a running tracker, Droplit would provide a single API for you to connect different types of smart watches for data collection within your app.
Edge API: Edge IQ
Founded in 2012. Raised $3.69 since 2014 from Diebold, Cambridge Silicon Radio, and Xchanging. CEO: Michael Campbell, previously SensorLogic, Innerwireless, and PanGo Networks.
Operator of an application programming interface and service exchange platform based in Boston, Massachusetts. The company's platform, enables organizations to handle data-intensive and device-centric computing functions at the edge of the network and simplifies integration with virtually any local application or cloud service through RESTful APIs and standard communications protocols. Practically, they are offering last mile management of connected devices: fleet management, provisioning, monitoring, firmware and OS configuration management on devices, distributed user and account management. They also extend the platform to handle ingest, normalization, analysis, filtering, and federating data. All of these services are exposed as a REST API.
Edge API: Seam
Founded in 2020. Raised $150K from Soma Capital and YCombinator. CEOs are Dawn Ho previously Sonder, Solidity, Plethora, and Sauce Labs and Sylvain Bohy previously at Instamotor and Nest Labs.
Developer of the internet of things (IoT) platform intended to provide IoT access control solutions for smart buildings and homes. The company offers APIs and hardware hubs that enable users to unlock doors, set codes, summon elevators, and provide access across door locks brands & intercom vendor systems, enabling developers and businesses to connect and control a large number of IoT devices with a simple rest API. Provides a “Seam Gateway” that acts as a bridge between IoT protocols and their API. Focus on security with firmware pinned certificates, full disk encryption, and secure tunnels. Provides enterprise features around fleet management, single sign-on, power over Ethernet, and battery backups. Seam is largely focused on residential and multi-family and not yet in the commercial / industrial segment.
Edge API: Mapped
Founded in 2020. Raised $3M from Greycroft in 2020. CEO: Shaun Cooley, previously IOT CTO at Cisco.
Mapped accelerates IOT projects by providing a developer enablement layer over your hardware devices. They provide last-mile connectivity to all of your hardware devices through a simple hardware / software package, and then expose direct access to those devices through a Stripe-like, very simple API for software developers hosted within the Mapped SAAS environment. Where most IOT solutions are full stack systems tailored to the specific use case, Mapped instead tries to simplify access to devices by unlocking the data they hold, normalizing access through an API, and then activating developers (whether through their customers, or through 3rd party ISVs) to create different kinds of applications that make use of those devices (monitoring, control, etc.). Initial focus is commercial real estate and monetize by selling the software that provides last mile connectivity, and also for API calls made.
Edge API: Sirqul
Founded in 2012. Raised $11.6M from Miteno and Owen Van Natta. CEO: Robert Frederick, previously Gripwire and Snapvine.
Sirqul is an IoT platform with over 70 modular services, which they call ingredients, to create a building block approach to building an IoT platform. They also provide reusable application functionality and dashboards to build different types of IoT applications that engage end users around gaming, communications and eCommerce. Once deployed, the system is accessible through REST APIs.
Founded in 2017. Raised an unknown amount of money from OSISoft. CEO: Thomas Arthur, previously ScaleDB, OptTown, RedSeal Systems, PhoneSpots, Arbor Networks, and Novell.
Developer of open source platform designed to simplify IoT data. The company's platform uses a modular microservices architecture including sensor data collection, storage, processing and forwarding to historians, enterprise systems and cloud-based services, enabling users to build better IoT edge systems easily, faster and at cheaper rates. Powered by Fledge, the Linux Foundation’s open source industrial IoT project which was originally sourced from Google, OSIsoft, and some industrial companies.
IOT: Foghorn (a DTC Portfolio Company!!)
Founded 2014. Raised $72.85M from LS Holdings, Saudi Aramco Energy Ventures, Intel Capital, Honeywell Ventures, Dell Technologies Capital, GE Ventures, Robert Bosch Venture Capital, Darling Ventures, March Capital Partners and Forte Ventures. CEO: David King, previously AirTight Networks, Proxim, Vitalink Communications, and McKinsey & Company.
Developer of an edge intelligence software designed to deliver the power of real-time industrial-grade analytics to resource-constrained edge devices. The company's software augments edge computing with machine learning to bring intelligence to industrial IoT which works with mainstream IoT platforms in the public cloud and can be easily integrated with AWS and Azure, thereby enabling businesses to make profitable analytics-based decisions. Where many IoT vendors focus on lifecycle management, Foghorn is taking the unique approach by optimizing delivery around complex and high bandwidth data in resource limited environments, which includes streaming ingest, data enrichment, complex event processing, edge machine learning, and insight generation.
Founded in 2010 as a Sony company. Raised $39M. Difficult to assess what their product efforts are working towards.
Founded in 2015. Raised $15M from CincyTech and TechNexus Venture Collaborative. CEO: Charlie Key, previously an engineer at a app dev consulting company.
They are an IOT-specific edge deployment platform. They enable you to deploy logic to edge devices, pre-filter data that will be sent to the cloud, create multi-tenant systems that isolate devices and users, incorporate a custom logic through a visual workflow engine, and data visualizations for streaming data from edge devices.
Founded in 2011. Raised at least $5M from Delta Partners, Investec Ventures, and undisclosed providers over a decade. CEO: Paul Glynn, previously Symbiote Networks, Fluke Networks, and Crannog Software.
A broad IoT integration platform that is built using microservices to deploy in different environments. Key value is out of the box integrations with a large selection of narrowband technologies such as LoRa, Sigfox, NB-IoT, LTE-M and then provide a low code interface for managing those integrations. These systems are then packaged as open REST APIs with a rich catalog for different types of management responsibilities including logs, digital twins, device automation, GPS location data, geo-fencing, and tenancy management. One of the few IoT vendors that discuss native CI/CD/IDE integration with underlying support for git.
IOT: Litmus Automation
Founded in 2014. Raised $10.2M from Mitsubishi and other unnamed sources. CEO: Vatsal Shah, previously Rockwell Automation and Arvind Mills.
Litmus is an industrial edge computing platform which provides data collection, analytics, and management in a platform that enables custom applications to also run on the edge through container deployments. This allows edge applications to capture edge data and transform into actionable intelligence. They also provide an edge orchestration manager which provides a centralized control plane for managing devices, data, and applications running on the edge along with their interconnects to systems and public cloud services. This includes embedded messaging for data integration, Dockerizing digital twins, Dockerized application updates, and fleet management.
Founded in 2012. Raised $7.1M from TVS Motor, Wipro Ventures, Lumis Partners, Pi Ventures, Infuse Ventures and The Hive in 2019. CEO: Vinay Nathan, previously Persistent Systems.
Altizon publishes the Dataonis platform which takes a big data orientation to industrial data. It contains an edge component for streaming data from OT and IT systems, an in-the-cloud data lake for storage and processing, out-of-the-box KPIs and dashboards, and a customizable business intelligence framework. Datonis is focused on pulling data into central systems, vs. other vendors that are increasingly looking at how to push computation and data onto the edge as well.
Founded 2015. Raised $22.34M from March Capital Partners, Honeywell Ventures, JC2 Ventures, Hanna Ventures, GE Ventures, Hack VC, Flight Ventures, and Juniper Ventures. CEO: Dhawal Tyagi, previously Aruba, Cisco, Brocade, and Cylink.
Iotium offers a network as a service bridge that connect assets over any protocol to any application. Provisioning is done in a zero touch techniques using certificates and key-based authentication without truck rolls or modifying existing security infrastructure. This overlay network is then layered with an IAM layer called OT-ACCESS that can provision, authenticate, manage and audit third party access to industrial systems. They also provide OT-EDGE, a mechanism to remotely administer Kubernetes running on the edge and to provide deployment of applications into those environments.
Companies Discovered After Publication (Not Yet Profiled)
Edge Cache: ReadySet.io
A caching layer that rides on top of MySQL or PostgreSQL with edge distribution.
Well, very forward thinking!~
Thank you. This is very enlightening.