Get best of the week

Autonomous Driving - hype in cars, but the need for data center network services

Over the past decade, data centers have taken the path of rapid development in the field of technologies and methods for deploying network services.

Not all decisions along the way were obvious or simple.

We want to share a scenario of applying the concept of "Autonomously Managed Networks" in the data center networks



What is Autonomous Driving for Data Center


The development of data center networks can be divided into three stages:



DC 1.0


The first stage, which is dominated by the simple consolidation of network data center data centers - we use the traditional architecture: STP + VLAN

DC 2.0

The second stage seriously improves the capabilities of DC 1.0 in terms of improving resource sharing, elasticity of their use through resource virtualization and dynamic service orchestration. At this point, the networks transform into a fully connected overlay architecture.

When switching from DC 1.0 to DC 2.0, we get an advantage in modern mature cloud computing scenarios and large-scale implementations of virtualized computing power.

DC 3.0

The third stage is aimed at adapting to the explosive growth of types, volumes and computing services in the era of Artificial Intelligence, and more specifically, Machine and Deep Learning.

The stage is seriously different from the previous ones:

  • superloaded distributed computing at several Data Processing Centers and peripheral computing nodes;
  • higher intelligence requirements for network architecture with deep integration of new technologies such as containers and remote direct memory access (RDMA) in modern applications.

Corporate and commercial data centers are constantly looking for "themselves" to maintain the explosive pace of development of cloud services. The desire to provide the necessary quality of services in accordance with the current requirements for “openness, capacity, scalability, controlled costs, security and stability” is becoming the cornerstone in the pyramid of values.

Classical tools for the operation and management of the data center do not show a visual model of the data center in the current growth rate of services, which depresses both users and service owners.

Against this background, regulators and consumers in the industry expressed the view that a set of highly intelligent and simple network management solutions should be included in the list of priorities.

What is stopping us, or rather, what we should consider:

  1. — , . — , . — 8 . — .
  2. . — , « ». — . , .
  3. Failures or errors. Time is money. Especially downtime for banks and financial institutions in general. Recovery should take seconds, not minutes, or even less hours. What does not always happen in classic networks.

How to be?

First of all, to automate standard manual procedures, to use automation technologies. But not all tools or approaches are equally useful.

Let's analyze the scenario of starting the network in manual and automated mode:



  1. In manual, we need to go through our template path - it will take a noticeable amount of time.
  2. In the automated, we are able to seriously optimize template operations, which already frees up time for creative rather than template tasks!

What's next? Let's see an example of the implementation of this approach:


But you say - let me, where is Autonomy, or at least Intent? The answer is already near.
An Intent-Based Network is what many have been marketing for the past few years.
The end user sets the parameters for the application, and already the network forms all the other necessary conditions.

Consider the example of data center services:


Not bad - but where is the Autonomy? Where is the effect of its application?

The whole system changes from passive execution to decision making based on recommendations. After the user enters the intention, the ADN system intelligently recommends the best solution. After the user confirms the recommendation, the system automatically executes it.

For instance,

  • , POD 1000 , ;
  • , , , ;
  • . , ;
  • 247 , , .



The data center network will no longer be a cold machine, and the entire system will become an active tool - an assistant - a thinker - an "almost" engineer.

Summing up - the key indicators of the ADN network - these are, first of all, recommendations based on intentions, verification of the solution before its launch and proactive work with potential problems.

ADN networks at the beginning of our development - we are now at the stage of - the transition from the “machine-to-help” model to the “man-to-help” model.



Looking optimistically into the future, we expect that ADN networks will continue to develop in five years - we will see networks with full autonomy.

What they will be we still do not fully know, but it will be interesting - we can promise this!

Conclusion


We bring artificial intelligence to our network solutions in order to develop the best solutions and methods for working with it. By putting AI at the service of “network autopilot,” Huawei seeks to reduce the complexity of O&M systems and failure prevention mechanisms to increase overall network stability. We really want you to like the way we do it.

More articles:


All Articles