Computing on the Edge or in the Cloud?

The era of big data. A very popular buzz word. Standing for the way how everything is generating and running on tons of data nowadays. Amounts of data that leave no other option but to build huge data centers to process them. Or is there another way? Have you ever wondered where data is actually stored and processed? And how does it work that you can access all that data from your smartphone? This blog post tries to answer questions like these by introducing the concepts of cloud computing and edge computing. Two more very popular buzz words.

It should come as no surprise that this blog post is not going to fully explain how the internet works. For simplicity’ sake, let’s imagine the internet as a distributed network of nodes where every node is a computer.

Clients and Server

The entire point of having an internet is that you can connect a device to it and download or upload data. That is actually all you do whenever you use the internet. Downloading and uploading bits and bytes.

Traditionally, the process of downloading and uploading is called the client-server-model. Where both client and server are two nodes at different locations in the network. The client is requesting data, for example a simple website request. The server serves the request and sends the requested data to the client. However, likely, there is no direct connection between client and server. Hence, the communication is routed through other nodes in the network. The nodes that lie on the route are called hop-routers. For now, let’s not worry how the route is determined.

Network Delay

When a client makes a request, the reply does not arrive instantly. Quite the opposite, it sometimes takes longer than you would like. A couple of reasons can be identified that cause this “delay”.

• Processing delay: Data is sent in packets between nodes and nodes usually have multiple links / connections to other nodes. When a hop router receives a packet, it has to decide to which link it should forward the packet. To do this, the router has to read some information that is transmitted with the packet. The time this takes is called processing delay.
• Queuing delay: A hop router can only forward a limited number of packets at the same time. Should packets arrive faster than they can be forwarded, they have to wait in a queue. The time a packet spends waiting is called queuing delay.
• Propagation delay: It is generally known that there is no such thing as infinite speed. Thus, no matter how data is transmitted, it needs some time to cover the physical distance between two nodes. This is called the propagation delay.
• Lost packets: In reality, packets are lost all the time, for example through an overflow in a waiting queue. Lost packets have to be retransmitted, causing further delay.

We will discuss later why delay is important to consider for both cloud and edge computing.

The Cloud

Cloud computing, arguably, has become the primary way to process data for many businesses. No wonder, considering the promises cloud makes. Virtually unlimited IT infrastructure, instantly available on a pay-as-you-go bases.

In plain English, you can rent out storage, processing power or software and only pay for exactly what you need. The only thing you have to take care of yourself, is that you have a powerful enough internet connection to work with the cloud.

Obviously, cloud can have enormous benefits for a business. It saves upfront infrastructure costs, provides easy scalability and enables businesses to deploy their systems fast. Also, maintenance and error handling are the cloud providers responsibility.

Literature typically distinguishes three service models.

Infrastructure as a Service: IaaS is the supply of network infrastructure, servers, storage, virtual machines and so on. Think of it as the supply of physical computing hardware.

Platform as a Service:?PaaS is the next higher service model. It provides an operating system so to say. This creates a convenient environment for the user to easily deploy their programs onto the infrastructure.

Software as a Service:?Lastly, SaaS allows the user to make use of software, normally through a user-friendly interface.

Data Centers

By talking about cloud computing, you inevitably talk about data centers at some point. So, what is the link between them? You can simply define a data center as the facilities that house server or storage and the hardware itself. Data centers show best the centralized approach of cloud computing. Gathering vast amounts of processing power in few locations and serving all client requests from these locations. This centralized approach brings some advantages. For example the concept known as automated failover. When a server crashes, the applications currently running automatically switch to a different server and keep running. Also, intuitively, things become cheaper in quantity. Which is very true for data center hardware.

Many companies have their own private data center, a so-called on-premises data center. Cloud computing on the other hand happens off-premises. That means a company rents out their data center and thus becomes a cloud service provider — CSP. A “Hybrid-Cloud” combines on- and off-premises infrastructure.

To summarize the cloud computing paradigm, in the above graphic the “Data Consumer” is the client and the cloud is the server. The essential assumption is that the data comes from a different location (“Data Producer”).

The Edge

What is the network edge? Different definitions exist depending on the context. For the purpose of edge computing we count all nodes that are “close” to access devices as part of the edge. An access device is every node that is used to access the network, meaning access devices are the source of a client request. So, for example your LAN and WAN routers are considered part of the edge — or edge devices / edge nodes — but not a data center or a router “deeper” in the network. Of course, “close” and “deep” are relative terms, so an exact delimitation of the network edge depends on the context.

As a broad definition you could say, edge computing is all computing outside the cloud happening at the edge of the network. One important reason for edge computing is the observation that the “data consumer” increasingly becomes the “data producer” at the same time. This creates a new paradigm! In a way, the notion of edge computing is to bring computing closer to the data source.

The Internet of Things

Certainly, the Internet of Things contributes strongly to this paradigm shift. IoT is yet another big buzz word in future technology talks. Not for nothing, since the number of IoT devices will continue to grow exponentially. IoT adds a whole new category of edge devices. Sensors, actuators and other IoT devices gather data all the time. In many cases, it makes little sense to process this data in a distant data center. Instead, processing it “at the edge”, close to its source is a more convenient way.

But why do we need the edge if we have the cloud?

Maybe you have heard how the 5G network is going to be so fast that everything will be available instantly over the internet. So, why not simply use cloud computing for everything in the future? Well, I am sorry to disappoint you. No matter what 5G vendors promise, do not count on the cloud too much! It is true that network speed and bandwidth will increase significantly but so does the amount of data. This is where network delays are the critical factor. For many applications, even a delay of a few milliseconds can pose a huge problem. Think about autonomous vehicles, airplanes (generate up to 5 GB data per second) or automated trading machines. Neither 5G nor 6G will provide the capacity for this type of real-time processing via a cloud infrastructure.

As explained before, edge computing brings processing closer to the data source. Logically, fewer hop routers are on the way. Fewer hop routers translate to less delay and ultimately to faster processing results.

Besides the problem of delays, other factors can make edge computing the better alternative: Uploading and downloading can be very energy costly. Data protection is often better ensured if the data travels less across networks. Also, edge computing can help to reduce network congestion and improve reliability of a service.

How will Cloud and Edge get along in the Future?

Having read this blog post so far, you might feel like cloud and edge computing are opposing concepts. And they are, as far as the fundamentals are concerned. However, their use cases differ a lot! They are not interchangeable technologies. Instead, the goal will be to build systems that benefit from both of them.

To put it in a nutshell, edge computing will be used to process time-sensitive data. Cloud computing will mainly process data that needs vast amounts of processing power but is less time-sensitive. Another way to put it: edge computing is useful to process real-time data, whereas cloud computing is the better choice to work with big data.

These fundamental differences have to be considered when designing the networks of the future. Along with the consideration of network delays.

Impact on our life

There is usually not much talk about advances in network technology. It mostly facilitates applications of other technologies. The list of exciting technologies is long, but their existence alone does not change much for the average person. In many cases, network technology is what makes it possible to actually use the potential of?Artificial Intelligence,?Virtual Reality, The Internet of Things and all this other great new tech.

Therefore, the impact network technology has on our life is huge! Just look through your phone and think about what type of data the applications need that you use. You probably find messenger services, apps that show you directions, social networking apps, streaming platforms, car-sharing apps and so forth. Even if the underlying technology existed, without the recent, rapid development in network technology none of them would be on your phone today.

Without doubt, cloud computing and edge computing will continue to play this role of facilitating all sorts of applications. A role that makes them key technologies for the future.

Saddam Hossain

Date: 15 July, 2023