Cloud and edge

Machine, enterprise and the amount of data generated on the edge of mankind exponentially.Data is, therefore, has a gravitational pull, and considering the factors such as cost, performance, and privacy, move huge amounts of data sets to a central location for processing is becoming more and more difficult to implement.We are into a world of data calculation, in this world, and its moving data to a central place, computing capacity sufficient than calculation will be moved to the near data generation, it is the ideal, is also possible.This kind of phenomenon is accelerating edge computing era.

Edge of computing to promote the industrial revolution and bring huge opportunities for the fourth time

The Linux foundation released the edge analysis report of 2021[I]Said: "as a natural extension of cloud computing, edge cloud structure is more and more regarded as" the fourth industrial revolution "the key driving factor, in this revolution, wide deployment of Internet of things (IoT), global sharing economy and the increase of the zero marginal cost of manufacturing, has brought the unprecedented drive by communication opportunity, derived a lot of economies of scale."

There are four to promote enterprises to enter the edge of the factors, including:

1. Reduce the delay
2. To optimize the bandwidth utilization
3. Offline or independent operation requirements
4. Comply with local (for example, the provinces and cities or countries) of regulatory or safety standards.

From games and digital content delivery toArtificial intelligence (AI)And augmented reality or virtual reality (AR/VR), many industries are affected by one or more of the above factors.

Edge of the basis of the calculation

The Internet the most diversity.Therefore, according to your concern, "edge" also have a lot of kinds, them as the edge of a hierarchical structure is helpful.Hierarchy of the various levels for round-trip delay, computing capacity from the terminal/storage capacity of the size of the total number of edges and a certain level.Equipment, for example, edge millions, distal edge edge of hundreds of thousands, macro, while public clouds dozens.It is worth noting that not a single architecture template can meet all the edges of the calculation cases.And, margin calculation and more centralized activities together in a public cloud, so the edge of the computing and cloud computing should be regarded as complementary.

As shown in the figure below, the following terms will help you further research to support the definition of edge out of and into the "cloud" design mode of all kinds of cases to lay a foundation and development of technology.

• Edge of equipment:Including smartphones, smart cars, surveillance cameras.Now there are billions of edge devices, and they in computing power, the network connection and power supply/battery function ability is different.
• The distal edge:Is contains the edge of a small amount of computing power (2-5 frame), found in sports stadiums, cellular phone base stations, storage room, apartment basement, parking lot, etc.These edges are usually place in distance data generated from 1 to 5 ms round-trip delay (RTT).
• Micro edge:Refers to the edge of cloud area (such as AWS where, telecommunications center office and flow aggregation point.The edge position in distance data generated inside the round trip delay of 5-10 milliseconds.
• Macro edge:Refers to the interconnection hub, multi-tenant city level data center, cloud and local area.Usually these edges from the location data generated inside the round trip delay of 10-50 milliseconds.
• Cloud:Refers to the computing and storage capacity of large-scale wholesale large IaaS, SaaS and model data center.Usually these data center from the location data generated inside the round trip delay of 50 to 100 milliseconds.

Edge computing technology development is the key to promote the development of modern digital infrastructure

The following key technical progress in the field of collaborative work, are accelerating the development of edge computing architectures:

• Container technique.As container level virtualization (such as Docker container) and distributed container layout technology, such as Kubernetes, can now build distributed control plane, the calculation will be moved to the location near the data generation.So, don't need to move a large number of data sets to a centralized location for processing.
• More intensive processing capacity:Increasing GPU processing power (" huang's law ") and the storage density (for example, one rack unit of flash memory capacity is 1 PB), in the distal edge and the edge of the relatively small calculation possible on a large scale in the physical space., for example, there is no need for large data sets will be moved to the center of the cloud for artificial intelligence model training.Now, on the other hand, can be adopted in the edge joint type/distributed approach to training and model inference.
• The high speed5 gNetwork:With high bandwidth and low latency of the emergence of 5 g wireless network computing can now from the edge of the device moved to the edge of the edge of the other types of hierarchy.Now, because of the heavy calculation work can be transferred to near the edge of the position, so can save battery power at the edge of the device.For mobile application developers, new software development model is used to make full use of the 5 g network is very important.
• Safe computing technology:With to edge mobile computing, algorithms providers increasingly wants to enhance its algorithm (their secret weapon) privacy.Also, the owner of the edge data would not through the public Internet to transfer the raw data or local insight to a central location for polymerization.With the federal study and other security mechanisms (differential privacy, homomorphic encryption, secure multi-party computation), the emergence of today could safely processing data, in the edge at the same time, according to data provider provider of data privacy concerns and the algorithm also gradually weakened.

What is into the edge/Move out of the cloud architecture?

Our President and CEO Charles Meyers told theCUBE[I]An interview to share this view appropriately, because enterprises in the exploration will move out of from the cloud computing and data processing and moving edge - according to the views of Charles, cloud or edge can be the core of the enterprise.

It is important to note that the edge architecture does not exist everywhere "all" solution.On the contrary, according to the delay, security/privacy, cost and availability requirements, a single application architecture can span edge of each hierarchy level.

In Equinix, we see the edge of the cloud out "and" into the edge "of architecture design pattern, considering the cost, performance, privacy and availability factor, is creating across the edge of the hierarchical distributed architecture.As shown in the figure below:

• The cloud movedMeans for performance (delay), data transmission cost, compliance/privacy and availability (disconnected operation mode), the application process from the center of the core/cloud migration to the edge.It is important to note that typically, part of the application will run in the center of cloud.With the advent of container technology, can now be calculated moved to the location of the data generated.According to the requirement of the cases, the data processing is on the edge of a hierarchy in the edge of the appropriate type.The cloud moved用例包括在边缘位置（也就是紧邻数据创建的地点）实现云服务，例如人工智能/机器学习、实时分析以及安全等。由于边缘数据的重要性，将计算移到邻近数据的位置（而不是云中）这一做法是合乎逻辑的。随着超大规模云服务提供商逐渐意识到，客户的许多数据必须存储在边缘，以满足其对延迟和安全的要求，他们正在开发本地解决方案，例如AWS Outposts, in the place of the generate workload run the workload of the first instance, and then a second copy on the cloud storage.Talk about AI, on the cloud model, training a AI first and then model reasoning model will be moved to the edge, it is very common.However, more and more application architects also hope that in the edge of model training, because companies are reluctant to transfer the original data to a central location for model training.We are now entering the era of AI, in this day and age, organizations on the basis of local data to build local AI model, and then the weights to a local model in a central location for aggregation, to create a global model.The AI will not only help to reduce the original data transfer to the cost of training center location model, but also help to protect the privacy of data will model the weight not only embodies the primitive data transmission to the center position.
• Enter the edgeRefers to enterprises for edge applications, server and gateway design and building architecture, which exist in the edge position application footprint.Usually, micro service and running in the cloud on the edge of the collaborative operation, and can be distributed control plane coordinate in the form of integration.Considering the performance, cost, security/privacy and usability, application hope on the edge of data processing.In marginal phenomenon, however, we note that, given the cost factors (delay), data aggregation, and performance, not its edge infrastructure deployment in every possible distal edge or micro edge location, but the edge of its infrastructure deployed in the city.For example, assume that a convenience store has 20 stores in a big city, they hope to deal with real time video data to monitor and improve consumer shopping experience (for example, according to the location of the shoppers in the store display digital coupons), in this case, if the video is sent to a public cloud, is unable to meet the needs of real-time response.Back to the example of the artificial intelligence, for cost reasons, companies do not want to install in every position in the city of artificial intelligence reasoning stack, but rather in a single location of the city to install a stack (the brink) artificial intelligence.Also, with the emergence of 5 g, without having to perform calculations in each terminal equipment, and can move up hierarchy enters (edge), in the center of the micro data for processing, because doing so can still meet the required strict delay requirements.

inPlatform Equinix ®On the edge

Now, for the enterprise, the deployment of the workload in the proper position, connect to the right partner, and coordination of automation, control and security is crucial to make full use of the maximum likelihood.Equinix can provide a for creating and interconnect global platform of digital infrastructure, to provide enterprises with competitive advantages.For example, through the global Equinix International Business Exchange ™ (IBX ®) data centerPlatform EquinixThe edge of the city location deployment of digital infrastructure, enterprise could mass closer to the edge of service and service providers.Including accessEquinix Metal™, this is an automated and interconnection model as a service (BMaaS), based on a cloud companies can go a cloud of environment and enter the edges, like the as easily according to the need to expand its public cloud computing and storage resources.

Equinix Fabric ™For Platform Equinix on more than 10000 enterprises to provide seamless software definition interconnection services, andThe Network Edge,Provides a virtual network and the security services, help enterprises to realize modernization of its network, and in a few minutes by way of virtual digital infrastructure deployment on the edge.These solutions allow customers to between the edge of the city and public cloud deployment of the edge of the agile and flexible service and access to, and no additional capital spending.

As a global digital infrastructure providers, Equinix with data centers all over the world to assist enterprises to enter more edge of the city area, so that the digital leaders could on-demand creation and interconnection based on digital infrastructure building blocks.Equinix make digital leaders, especiallyCloud native enterprises, and can be trusted platform to build their own global edge.Through Equinix Metal using the computing and storage infrastructure as a service, through the Network Edge deployment in the OPEX model and virtual Network security, save the deployment time and cost.Moreover, Equinix Fabric can be combined to our global IBX seamless connection of all the products together.

By digital infrastructure (network, computing and storage) placed in the edge of the adjacent public clouds, the company can be deployed out of/in the edge of the architecture of cloud, easily and economically solid shortening application delay, improve performance, reduce costs, and enhance data privacy/compliance.

For more information, please readThe Platform Equinix vision report".

[1]The marginal analysis report of 2021, the Linux Foundation, 2021