What is the Distributed Cloud? What are its main principles?
Knowing how to manage its IT infrastructures with the best performance has become one of the main challenges for companies to gain competitiveness in their markets. The services of Cloud Providers such as AWS or Azure are numerous and the objective is now to bring them closer to their users.
To meet these expectations and in line with Edge Computing, the notion of Distributed Cloud officially entered the vocabulary of Gartner in 2020, which listed it as the main trend in 2021 and called it the “future of the cloud”.
Also read: Top big data trends of 2021
Towards 100% availability and optimal performance of IT infrastructures
Since the advent of the Cloud, Cloud providers have been looking for the best possible performance from their infrastructures to meet the ever more demanding customer demand for real-time access to their favorite applications. While this research may prove to be ambitious, it is a key element in order not to generate a loss of turnover or an image deficit.
Several factors are the cause of downtime such as failure, failure of data center equipment, an unexpected weather event, human error following repair or maintenance of equipment.
To overcome this, companies, therefore, have several Cloud strategies that can be implemented:
- The Private Cloud : dedicated to a single client, the servers can be hosted internally or by an external supplier
- The Public Cloud : deployed by a third-party cloud provider, it is therefore open to the public and shared by users
- The Hybrid Cloud : it is a mixture of private and public clouds linked together
- Edge Computing : Public Cloud functionalities directly on the customer’s physical site
Although Cloud Provider data centers are optimized and secure, they remain far from the customer’s physical sites and limited in terms of performance for any company wishing to keep a hybrid approach. Businesses now want to leverage cloud platforms and take advantage of public cloud services wherever they want, not just within the boundaries of traditional areas or regions of public cloud providers.
The main principles of the distributed cloud
Distributed Cloud may seem very similar to Edge Computing, it is in fact its evolution. Both technologies indeed involve processing data as close as possible to where it is collected. Unlike Edge Computing, Distributed Cloud will make it possible to distribute public cloud services to different specific and varied physical locations (on-prem, Edge, or on different Public Cloud providers), while governing and operating them from a single public cloud provider. The Distributed Cloud will then use edge solutions and gateways with a local network and act as a bridge, bringing together the cloud and the IT systems that are at the edge of the network.
Very concretely, these are AWS or Azure servers that are directly integrated on the client’s physical site in which we find the desired services and functionalities of the Public Cloud and which remain managed by the Cloud Providers. The ownership, operation, updates, and evolution of the services will be the responsibility of the Cloud Provider. As if the most important features in AWS data centers were directly accessible at the customer’s physical site.
For example, AWS Outposts offers AWS machines (the same as those in AWS data centers) that host the AWS services that the customer wants. The goal is to benefit from both the power and the ease of use of managed services while minimizing latency.
From a network perspective, cloud services are located in local or semi-local subnets which allows them to operate independently and intermittently without being tethered. The data is thus transmitted from the collection point to a gateway to be processed there, then sent back to the periphery.
2 scenarios are then possible:
- For data requiring high reactivity : processing at the local level by an intermediate machine
- For less sensitive data : sending to the Cloud for historical analysis, big data processing and long-term storage with the aim of building a predictive model
What are the benefits and limitations?
In addition to the fact that these are the same advantages as the Public Cloud, namely exploitation of the hardware and software infrastructure by the Cloud provider, a financial and technical elasticity of the Cloud and the opportunity to benefit from the pace of innovation of different cloud providers, the Distributed Cloud makes it possible to:
- Improve performance in terms of latency and availability. There are therefore fewer network incidents and a reduction in the risk of overall network-related failures or ineffective control plan.
- Optimize data center infrastructure costs by reducing the costs of running on-premises workloads.
- Solve the problem of sovereignty by being in accordance with local legislation, thanks to local data hosting.
- Be flexible on the hosting strategy. Some local services will be processed locally and other applications directly migrated to Cloud Provider data centers.
- Accelerate on innovation and adoption of technologies by development teams. They have the possibility to access the Cloud Provider universes and to use new offers and functionalities. It is now possible to quickly test new architectures to validate their designs before quickly implementing them in the production environment.
- Modernize legacy applications using any combination of public cloud services.
Although the Distributed Cloud model has many benefits, there are, however, three major limitations:
- The Complexity
Distributed computing systems are more difficult to deploy, maintain, and troubleshoot / debug than their centralized counterparts. The increased complexity is not limited to hardware, as distributed systems also require software that can handle security and communications.
- Higher initial cost
The cost of deploying a distribution is higher than that of a single system. The increased processing load due to calculations and additional information exchange also drives up the overall cost.
- Security issues
Access to data can be controlled quite easily in a centralized computer system, but it is not easy to manage the security of distributed systems. Not only must the network itself be secure, but users must also control the data being replicated across multiple sites.
What is the future ?
For Gartner, the Distributed Cloud will evolve in two distinct phases. His vision is that companies will buy cloud substations to avoid latency issues. There will therefore be a dramatic increase in the number and availability of places where cloud services can be hosted or from which they can be consumed.
At first glance, customers will not accept the idea of opening their substations to their close neighbors and will preserve the use of it in their premises. The next generation of cloud computing will work on the assumption that the cloud of substations will be located everywhere, like the Wi-Fi hotspots. The physical boundaries of the cloud will therefore be broken, which will allow companies to overcome latency constraints and open up new possibilities for reaching customers in dispersed environments.