Edge or cloud?
Edge computing and Cloud computing have fundamentally different functionality. They are usually discussed as different approaches to compute infrastructure. In actuality, they complement each other and work well in tandem. Today, we are going to talk about the benefits of Cloud vs Edge computing, as well as the hybrid Cloud/Edge approach.
Our Chrysalis Cloud platform is for securely and simply streaming (real time) video from edge devices (ie IP cameras) to the cloud, easily and inexpensively. In the progressively increasing real-time AI world, we are in constant talk with our customers about the trade offs between computing on the Edge versus Cloud computing. As AI becomes increasingly ubiquitous, more power and more resources will be required to run and scale projects.
First, let’s dive into the background of the computing market. The markets for Edge and Cloud computing are booming as IoT devices become widely (and cheaply) available for consumers. There is a need for analysis of data generated from the wide variety of sources. According to one estimate, all people, machines, and things are expected to generate a staggering 800 zettabytes of data by the end of 2020(source).
–> The global Edge computing market is anticipated to reach a mere $43.4 billion by 2027 (source). Companies including NVIDIA, Google and Intel are developing processors specifically designed for Edge computing technology to accelerate the inferencing process for AI and ML applications. For example, NVIDIA’s popular Jetson chips have made waves for exceptional speed and power-efficiency in an embedded AI computing device.
–> The global Cloud computing market size was valued at $266 billion in 2019 and is expected to expand at 13.9% a year from 2020 to 2027 reaching a colossal value of $718 billion USD (source). The cloud infrastructure-as-a-service (iaas) has been largely provided by Amazon Web Services(AWS), Microsoft Azure, and Google Cloud Platform. New technologies in AI and ML have tested the capabilities of those companies and has opened the field up to other players. IBM, Alibaba Cloud, Tecnet Cloud are some of the new companies in the market.
Chrysalis Cloud has now entered the market with its superior cloud streaming technology, custom-built for streaming video in real-time AI functionality.
Now, I’ll outline the details of each solution type for processing video data.
Edge computing is the process of capturing and processing data close to the source, either in a device or via an on-premise server.
There are a few advantages to this method of data processing. The primary advantage is reduction in latency, since the compute power is readily available on the data generating device, the time spent exchanging data for processing is negligible. Edge computing is also often seen as more secure, since it reduces the risk of data exposure across multiple devices. When data is processed strictly on the edge, it’s possible to build a private architecture thus increasing the security of the data.
The main drawback of strictly Edge computing stems from hardware limitations. Updating hardware at the edge, finding bugs and fixing them is time consuming and inefficient. Storage and compute becomes a huge issue when dealing with solving problems involved in AI, data science and ML. Further, only locally accumulated data is available for processing, making it difficult to run any sort of advanced analytics.
Cloud computing occurs when data is gathered and transmitted to a data center for processing. The transmitted data is also stored in the central server. All devices that need to access this data or use applications associated with it must first connect to this server. The devices need to be connected to the internet whenever there’s any need for transmission, processing, storage or retrieval of data.
The centralized nature of cloud computing means that processing time may be slower for devices that are not constantly connected to the internet or that have a lower bandwidth connection. The bandwidth costs involved in data exchange can also be high, which drives up the overall cost of computing mainly due to current on the market solutions.
Cloud computing is far superior in terms of power and capacity. There is a vast amount and variety of CPU/GPU computing power available in data centers which can be accessed and used on demand. This scalability feature is a huge benefit of cloud computing over edge computing.
Cloud computing also allows for gathering massive amounts of data and analyzing it in a variety of ways. This capability has allowed fields such as artificial intelligence, machine learning and computer vision to become viable business tools.
Drivers for Cloud computing include demand for IoT sensors and CCTV cameras that require analysis of large volumes of data from multiple sources to gain real-time insights.
Hybrid Approach & Chrysalis Cloud
By incorporating both approaches, a powerful and cost effective system can be built – maximizing the potential of both systems while minimizing potential limitations. This hybrid approach is becoming increasingly common.
AI and CV applications require processing large amounts of data quickly. This data needs to be stored in a secure and structured way. Further, models need to be retrained constantly, which requires sending data to the cloud for processing centrally. Relying on a purely edge solution for AI and CV applications would be difficult and costly.
Given the number of potential use cases processing streaming video, creating a functional future-proof architecture is key. Use cases include any instance where surveillance from a remote camera is happening: in-store retail intelligence, real-time facial recognition and security, medical monitoring, object and brand detection, robotics, etc.
At Chrysalis Cloud, we have created a set of solutions that allow our customers to take a hybrid approach OR a purely cloud approach.
The innovative edge solution from Chrysalis Cloud- the “Chrysalis Cloud Edge-AI Proxy”- allows users to ingest multiple video streams and has a common interface for conducting AI operations on or near the Edge. This an easy-to-use collection mechanism from multiple cameras onto a single more powerful computer. For example, a network of CCTV RTSP enabled cameras can be accessed through a simple GRPC interface, where machine learning algorithms can do various computer vision tasks. Furthermore, interesting footage can be annotated, selectively streamed and stored through a simple API for later analysis, computer vision tasks in the cloud or enriching the machine learning training samples.
The proprietary streaming architecture from Chrysalis Cloud, the “Chrysalis Cloud SDK”, is specifically designed for cloud. This SDK provides easy and powerful control over live media streaming consumption and ingestion into various machine learning libraries in the cloud. This platform has sub second latency when streaming to the cloud (compared to similar systems like AWS Kinesis) and at a fraction of the cost. We have developed an innovative easy-to-use collection mechanism from multiple cameras onto a single more powerful computer at the edge, where machine learning algorithms can do various CV tasks.